Bestand wählen
Merken

Optimization (25.11.2010)

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
parallel of pleasure began to welcome Yalta the that a where housing that mining lectures and last week we were talking about the next talked well basic you were talking about in the structure of the UK could
be time but it turned out that my early next touches as we owe McGovern about it in a really so that the basis of on that allows you to full model dimensional that on the ICC to especially Peachtree's load of 1 the middle type of the trees can be used and the best rose specially structures of that can be used for multi-dimensional better than we to the opt free which basically it is OMX Tree consisting of minimum balding rectangles and the minimum bombing rectangles and close without a points and some of the of structure key at I'm incensed that so that the structure and then we have discussed several that Ms of the opt freeze us all overlapping between neighbouring minimum bombing rectangles which may need to several about several each of the last out which doesn't allow overlapping than rather the multiple of versions of of the rectangle and finely the on the street that is 1 of the most Fleming off trees but we are also talking
about some of our and an extension of the law of the tree at the ubiquitous the treaty which is a space will be that are uses not slow left right decisions for Palm traversed in the notes but of rather spacelink about which could be the sacrifice and and and 2 of the 3 but could be every other spacelink the with of salt piano code of what every need a select cambium lamented and find LeWeb going into but met indexing where we saw they upgrade for tables was set like at 2 beautiful for example if you are male female that is just 2 but structure so you could say 0 mayor and 1 this female of efficiently to stall this and very efficient to to build filled on that to build operations on that because those will be because operating thought where which makes it very cost to use it made in best if you have a lot of domain values for network beauty a bit indexes to fish because of the things that is basically what we came out with all last lecturer and in this lecture we want to
talk about period of setting a little bit further in terms of optimizations so some of the time want to talk about is basically politician a halt to petition that into a helpful I'm area so that all of the of the few together should be politicians on to wants to avoid into 1 table table excessive and the kind of optimized than we will be talking about joined joined as 1 of the basic operation before for that based on the board of a program basically and we need to Paltrow and very strongly because and and allows you can easily imagine that the type of joined having such beat them in the face of the huge factors that is definitely 1 of for speeding the and the last thing we want to talk about it to a few of the questions about we have a lot of information that some all aggregated and the cubes do we spread just the that cells and and build or the aggregation's on during their time end we want to materialise solve the aggregation to gain that up to 4 months in recent dressing up of petitions solos aggregation's of this space and of calls which wants to choose the best stop was the partitioning of death
and the basic ideas that year you you do would have been 1 big multi-dimensional Cuba but in the that warehouse as a lot of debt and talking to her about him and the wild and stations at every solid on single this it was stall that on singles of and the answer to that of 50 no because we have the data for the scheme of the usually is very very very simple if this 1 big fact tables having all measures the many of the them engine a tables that local structure may depend on whether you to style snowflake scheme of but if as if you do stuff scheme than also by mention tables made before so do you stop them together and let for what do you do the party Titian them in some way off but the answer is that new cost told them altogether unique to define certain sensible politicians and
why would he do that because 1st managing that that is so much easier if you just working on the small petition containing just the real that preferred example conciliar yet updating some information the puck domain the inefficient to do that with a few have over data after look for all records the where the product information it if you break the news to see that the cocktail of the public domain straw in a petition way to only said this single-table I'm as well if you do restructuring what happens if you have a diamond land based mums and I saw something like that and some 1 of the application that interested in UK needs weeks we need and what I do you restructuring at tables of this if you do that on the big table the very inefficiency set to do and it will basic across the whole where house downtime because nobody can were up on the big beach tabled once you Lowry petitioning at all read ranging at om saying growth for the next and if I of small petitions you can and the more easily and also maintaining the and becomes of the inexpensive bomb if you use petitions you may also speed sequential scans because getting the full tabled obviously take much longer than scanning part of the table is the possible table well chosen that is a good thing for Sequential SC and of course also recovered Ian and wanted soaring if you if you have less items to look at set up in debt and the
houses we use petitioning to 1st improve through the for 4 months but the obvious way so we wanted to minimise the amount of debt that it really has to stand by will also interested and availability of the day petitioning that means that we can put the petitions on 2 different that may not only be physically but also geographically district cyclists over in America in Europe and Asia and the working in Parallel and use the downtime Sodomites times for example due to all the at Minnesota stuff organisation on 1 of the service because they might at some point in America says it all your an Owais and I'm which is well and good for full back up the hill with some some kind of summer Staum its the you S and Hugh Computing said the breaks down euro but the fact that if you petition that an replicated all the different geographic regions of the very good idea that are also embers if you need some recovery issues that much easier to do the smallest of all smaller data volumes that these items of 2nd saying is that in the station so as just pointed out the have at columns all restructured table all deny sometime in that much easier if you smaller that and only that by mansion will be down to wire I opened in all the other stuff amendment all the other than many of the fact they can still be guest by critics of the River point in India in the one day to politician of that you and now along 1 thing in his favour and amount goods staff cause them anyway stupidest new that and will want to talk about the following I'm about to different ways of their and 1 of the benefits of each way so that we could come up with coach for the sight of
a white for the great I'm in depth auditioning for a part dispute that allows the several approaches 1st as the that that petitioning which means that don't necessarily petition the hot where of the tables and put the damage that a full time in 1 table and the damage that full of the 2 graphic eyes them in Fife tables with respect the 5 Continental something akin to end the said you can petition by David or by a line of business the cell some businesses only needs that of from some some to graphic regional for some products that selling out geography under the organisation a unit that I'm concerned with so if you have Wales contains that a full foyer accounting I'm this could be published and although the holes at the where house I'm there any combination of the 2 so that can be very difficult to find out what is the best way to publish it on the other way by on the other side could use half petitioning which is basically making the data available for multiple Processing notes of fell this kind of Distributed database that you need full building the really big getaway houses anyway secret throughout the long table still 1 but table that it in fact distribute of different different of of different hot dry and them well then you have to make sure that all of the songs often although Hopcroft saw ready for work from the group and 1 of
the problems were problems that you have this about the level on which the petitioning should across you could either state basically it's an application questions that the battle with the also FFTW that should look at the crease that showed a kind of like a got what is the best petitioning schemes and and is responsible for so it knows well tables are Llandudno's where everything is I'm of cause of possibilities here for doing it in the right way with respect to the application was respectively reload suspect to the needs of the users on application of history a ban on Africa application of a have to care you to implement all the different excesses that you really need witches 30 off not what you would want to sell the 2nd possibility is just to let the database management system care for Disney and just go OK at on all where the table strong but more Mulhall the petition to ages was to create a sense to create database and that of a figures are all the excess P offers at what joins on needed and and what of those magically for me it's just 1 b table for the application of justice singled out table I'm if you do it on the database management level its at this kind of had a clear how it works that database take care of course the database which will do it in end sometimes unpredictable way so what the application does all was but of the game to go off off few business processes of this no effect whatsoever on how the database management petitions system published to that of the database management systems used some Algorithm for assigning the server so exciting of the hot dry for certain of the so things that also include all put on break cost devices all something like deserved a structure or the you estate where you will find the best man of Indian database management systems but he application knows that someone up losses are Berlin Pullman Ozil or Walker more often than not but it doesn't have any effect on the whole the database opticians without at the centre of the really make sense to do it I'm at at at Applications level of for example if you have an application but for a approval but the post tax and Mcshan tax-loss change every year the base of an oil nothing about tax rules applications does for petitioning the that in a sentence that are those that other belonged to the same tax paid taxation period before will change aloft law change may be a lot of sense semantically that the system will not be of and in every created you write if to specify these these intervals but if you do it on my application of made just dress write table and that of the 1st period after the text will change of and so this is basically a basically walked what can happen if you do it on application will but still with the 2 law Mohammed fication overall on the basement system-level some of year
so that you see the abuse of some sort to a are this petitioned by the notebook up the into several slots with a strained by the company and and for the moment in the year when the group of of of so
what their petitioning basically and lets is but you have to split the week table you might to them and we have won the table basic at of whether to possibilities displayed at 1st you can't good enough horizontal way of and from the them crew with can of still may be afraid to
say that it will help solve basically you either may split of horizontally or you may see the work has onto the means that different rows of put together and you create a keep or be at tribute for you Trow vertical means that although the at tribute values for single at tribute cost altogether or groups of tribute and you rose for this specific at reviewed but you distributed the at rebukes 1st petitioning schemes may make sense because for example that would be a good split if you come to the UK tax law example PM at exact books that would call for a horizontal split because you have to run the fundamental change and everything after that it is want what such a thing before the them on the wamu Peugot idea for Burczyk of split yet yet read it yet and this book could be interesting politicans but it Yell but would be a PoW tuned to pop Monday St meet the addresses of customers pay Seagate just interested in the purchase as the summit has done not who bought it from so wide left to have all the different at to be land to some purchase the number of items the different set off items the prices the date when the purchase was done by the time the child under about mobile the that the supply 0 delivered the stuff the address of the custom bought the stuff why should old be in 1 single-table if I of marketing just interested in the customer Web stuff and his her dress to send a Christmas card in the light of the new custom of sentence and advertisements to and a tiger specific group of customers in the other case the accounting to palmone would just look at the prices and the dates when something more sold and and also but never to customers such to split them into different tape and this is exactly what happens is that the horizontal petitioning you take a different rose and put them into different tables it is a vertical petitioning you distributed the abuse into different tables of this discuss the enough quest want to split the atom beauty is still need to know which will be launched to get out of it yet or O'Hea and this part because he and and this part the city and the need 1 identify Tenali these 2 rows together which basically is of calls the primary key sell you have to duplicate the Prime Ricky into every table has been voted the horizontal split you don't to cricket and the Rose of all complete UK different understood self this basic do what I do that stop was the horizontal testing where you do is you split the state of tapas among disjoint
table the every tackle Whitaker in just a single table but in all the tables likened the cause of it and the verdict if was a basically you to take a set off release which is the horizontal petitioning off some last of Relational last Erakat if every politician really is a part of the must attend the petitions side disjoint no will cross into petitioned and it is up would ever politicians to get the then this will result in the hole to a full lost in the petition update such easy to see why I'm so I'm if you have a different petitioning procedures so could to a different part and procedures are that might result in different for non petitioning schemes to you might budget by arranged with sampled the texts example would be arranged published every Reckitt before the 1st of generate 2 thousand and everything after 1st of the 2 of you might have list petitioning waited take so I'd take about the 100 customers that are big customers for my company and then those records that concern with these because the most applauded 1 table and the rest customers are put in some of the that might be a way that I could use hash petitioning the of some hash function everybody knows what a function as has not as good I'm so some hedge functions that just Calculate Ashwell you over every couple and depending on best volute its board and some paper which is kind of a step petitioning the advantage of pressure has petitioning as wild 1 do that because behind at front office is as no semantics likened the text or example of the best customers of richest customers that sample load you to hash petition if you exactly that as bad as the of the so abstractionist care you can or predict for both actually calculating the hash of about what the hash of each up couple would be off the record will be and the records of their for distributed even the Oval the table politicians and that is a good idea if you are concerned with distributed that you can be that the RA anointed and because of the work again not the way to today so that it
is obviously as the horizontal petitioned from which is basically very similar to those of horizontal petitioning only that it that you don't Distributed different coupled with not semantics of distributing your customers or distributing specific times of the euro something over different of those but making different at tributes of or you must to record accessible on 1 device and she before example marketing of need some of information in the accounting development this would be a good idea well it does that doesn't matter because they count seat questioned offer that at the root of the
PM but yet but I've learned a lot Skolnick yet could put a well yes Supreme up sorry for left delays but steps kind of which had Adam walk said she doing well see you turned the eye and said UK at As tried again this
time with as much and and the
duo talking about horizontal petitioning that can assault pressing
range petition is selected
petition by determining the key fell
you inside certain range and either
be bomb the primary
key there or can be some of of active you like the date of the type of ball of what every you are interested in are basically it's about restrictions on the mouth so you see elect from the mouth of table a number of records that is petition to certain value for some critic of the so it could for example gold country of Germany and to sell my country Germany and ideas smaller than to among all the other countries of the ability range petitioning of enough horizontally of a tape of St the this petitioning I'm you assigned list of values will the of computer of has become all which or you almost appalled customers older but you could just leave them and if the petitioning of while the values you choose at full petition you take it out of the mouth tables and if it doesn't you put some of table so for example you could all road take all roads where the country's Iceland more ways we of them out which is basically Scandinavia the information that the police country's all Scandinavia and no other country Scandinavia a semantic inflammation but you have to provide some of again problem it's a restrictions but this time of the selection but basically St basically a product that is in some list but next part of some of the goals of the compliment was not in St come last but is the hash petitioning Asiaweek told you as a hash function the 5th time the membership of each petitions so full by if you need for petition for example you need a hash function that returns by from of 0 2 3 still something moral for for example of for each couple from after tabled you Calculate the hash functions and depending on the value of their fund to return st could you put it into the right to publish by the useful if you need to show load evenly between you different devices how to choose the hedge function as is a difficult call them and sometimes to overhaul to chief but if you looking at it that way housing the horizontal petitioning very of walkers by the time the engine so if you put together the that of 4 different fiscal year is for example of the very off the full of the most recent data chills current strand all time trends forest of my that L of longer lasting data shows some global evaluation of the and the and the idea of if you take them and other than time than it basically grouping so for example you could look at different branches of the company he could look at different countries of the company and stuff and there is a very often than by this end of the table side is that they are for the most hash functions of the really big Taylor if you really have to care about them at the thought of that that it might be sensible to display compound of small pieces and if you don't have any nice key to distribute them by and just Hughes the hash function to get out pieces of about the same size but the other possibility is a vertical petitioned as ordered a told you is basically pudding the columns appalled footing the at tributes into different petitions I'm and
so basically that distributes every single couple of you must record over a certain number of people that if you are interested in a complete couple you will allways need to join table but if you Applications not interested multiple or at least not very often that and it might speed up the about Eastern time considerably so I'm glad I can do it is a bit basically comes down to rose between what I can do is to look at what the of needed about some item of some entity for for some of the custom of the for the public that all the countries of the total sales that also and the and the spot that you identify the very of next together in some serious the should be politicians out of the of the whole table and some your words after to take the I'm the Prime Ricky to identify which rose belonged to get it in the mouth of tabled sold as basically a one to one relationship between the 2 for petition every couple is evenly distributed to every petition I'm in their housing it means that you use their ready used columns in some different if you could put that table of the Royal used items are on to some slow devices that were genius hop where most companies will have full Pukach of or you don't just don't have to to protect it if you run period on the high used that so basically phase of protection include revaluation which definitely need some time see the that if you believe you across the different politicians than you end up with some of the the origin of the table from and of calls the Super some sympathy almost penalty because of the joint costs and drawings of 1 of the very cost the operations and that the houses still if you really very often need and Thai erected their a bad idea to petition perfectly so that is basically what to stay they with the verdict of what I'm if you have a very large dimension tables are so if you have tens of millions of customers all something like that then very smokers that some of the attributes of the demographics stuff if you buy something at the full Knollwood about the value of offer for your was a coach 's information that go somewhere that also inflammation of the Red used because of the accounting is not the interest mocking so I'm this is a typical kind of data that you would put into some of the dimensions of a low priority and world that is basically what you do in still you need to maintain a link between the fact table and complete customer demand basically this going from a start scheme of snowflakes scheme because you but several tables in 1 them a pan with the kind of penalisation that across fragmentation he the of 1 solution this very often dumb is a so called many dam which is a special case of of the dish and that is what it would I basically do is kind of building the snowflake scheme among those petitions of a tribute that long to get up and order obese dimensions of and kind of like the mean I damaged so if you have the customer demand and to customer has some some of the dresses and all that that others went was a customer but he may also some demographic that for example and sometimes you need to the customer the that customers and sometimes you need just demographic that behind many male or female customers what this fall you don't need to accept the specific customers but you just need to excess the PM demographic that so you basic you recommend the diamond of customers into 2 little time and we need a 1 is the cost among still and the other is the demographic effects of customers and of course there correlated although the as a one to one relationship between them because the every customer and the cost Matteo 1 set of democratic that if you completed and not training and that can be period independently of each other that is kind of like good idea for the next and so to make the example if he affect a book that has the customised the and and you have the customer demand Jim from the customer demand may also contains like lost
their 1st never dress the age group and the income Group and the gender of the custom all the area was zip load of stuff like so instead of quitting that he and vast making the customer table there rebate Ikon into the use a key that his direct part of the attack table and to basic the what that means is that the customer demand is broken down into 2 engines where as the 1 to whom relationship 1 2 under relationship between the record in both tables but those tableside directly addressable from the fact that that Mike and ask series about I'm what customer group that something was about to serve for the dust by using this collection him but I ask what customers old what is the name of cost must find some followed by using this relationship he is basically the ideal for many damaged a red and 1 of simple comes penised M the variables and the minute dimensions must still be disinclined and you have to have a key the as a fan key in the
factory and in the diamond table from which the mini dimension to city of a one on one relationship between the record in the men's table and minimum and the and you have one to one relationship 1 to n relationship between the PM at the entries of table and the media that the UK and some of it should look at the help menu dimensions don't get to be sell the Bali advantage of petition basically wreckage used to get a group together but if you do period and the curious to a
and it does not scanned the whole table it just scams the petition off the table fragment of take much more efficient each petition can be optimized security reason for cover region with a reason to do petitionary application all reputation of all that I'd you can spread the petition fund different this so eightieths Load assault after this so that her genius infrastructures like almost every company as it would be a good idea to put it on to different December of cost of the also red that of used a on Foster devices of ready used that the people of Saudi vices I'm end of cause you can take advantage of the of possessing capabilities so for example if you have a petition to the dead of the year and you want to have a year the average as you can at the same time it all the different politicians and Calculate the aggregate of the can be done totally out it on the other hand had a if you have to retrieve that across petitions you end up with the bands of joined the were to keep up to fish'n'chips to build the coupled with horizontally petitions here to do a union all the different results at the end of cost to the complexity of what you have to look after database at the fate of and you have to make sure that things which basically just 1st the which is practical petitioning example but the soloist in that what they shinning can help unbelievable petitioning
Guggenheim practical examples of the type of low taken example from Oracle Shia and were 1 of the 1st grand questions which comes in mind it is to win should we actually used by teaching Wembley but they should have to with and it seems that the only though make amends for their databases tools think about what they should be pulled over from petitioning when you have big was which on paying a large amount of data and to the ground by well they don't below give any reason for that saw what kind of suspected that its old people so is the ball to a the confirming the window operations sees them some more Poisson systems and to marry be 5 difficult Basle at this is just speculation speculation of mine highway where most of the event or say it will be a great idea of what shouldn't tables which are created and to gigabytes OK about what the Bible numbers if you want to know something like what is on the up but they shouldn't be and if you have more than 100 million goes down on fishing is the way to go and you should or a based on to a worrying about the need to have more than 1 million goals so this is kind of what you want the of the part practices that there us about 2 away by the shooting of cost this comes at a cost and it should not because they are being became so if you have a smaller they will you should of mobile petitioning just because it's called the petition what they were on the costs which petitioning comes with is 1 of the town's quiet and see the idea is that the database management system should be along with the petitioning the application itself so everything which happens biological level should have no influence over the idea that they would is population physical petitioning should be transplant applications and everything should be covered by by the database management system based funds that too is a supplies but not all the database management system so that it can buy you a baking sheet and an example from record of 2 sold just for problematic the city's in their case so that start with creating the party should you can create though kinds of politicians in local from range used edged up the system politicians and you can also combined where
it all starts with the creating the fabled classical create fabled and then not have the possibilities for the search and population by arranged to establish the established arranged by the UN he tools and the time I'd be sold time her by mentioned and I've said OK politician by a teenage by my the party should remain words before the thousand 8 and the CW strain which allows me to walk were on the phone but best is value is less than and I'm going to to was the sole of they expresses before the of the thousands everything but and that goes in PPL before the thousand the same poll patient put thousand date with the latest than and growing told give the 1st day of the thousand 9 so everything we choosing between will fully into this part if want to go any further than it website for got to see what I'm going to do with those of a new information everything which are moving from the which is not 1990 willing for more the fight with the
label of the newly created the cable and say OK I'm going for a constraint and say that the want into politician after the thousand paid which comprises videos outside the warranty created by patients but less than Max which is dynamical caused unpicks facing the posted today a for example so it it is a tool Belgrade on patients of the with something like this so he would have be flown thousand paid if I'd have this kind of the formation of political that will take you through
the she close physics of OK there have information from the thousand pages and them everything we choose about the thousand they found them to be is in the book by what should think it is pretty intuitive but will not comply problem from the beginning we said that the town's by is not that easy enjoy what if and that's a pause I've said that the
powerhouses we don't Book of public this a wedding that 80 opposes and were going off the recreated or petitions blue and white if 1 for the debate on the state and the city will have time stake is not from the thousand 7 basis wanted thousand can happen but what goes says Wolves we have a problem we have a problem because of the update something 1 which of what they should sold for my Applications this is not as but and any more because his as Mikheyev about this and that I want to modify something solely on about it because it would appoint but a more that as they say in the way we have to allow for the means to build something like is because if he is 1 of 11 new in a note his should be allowed for quite so obvious as well so that the state is in the local meant they shouldn't you would
find in a fashion which has enabled movement so allows you to move roles between part patients and then and it's OK you can do it by what you should know is that if you want big that record into the details from the politicians will population instead new 1 in the previous even though I'd be which again is not as bad and you now have a new I'd be which should be out but the database management system the by and why was it necessary widely duties because of cost him for example something like a primary index and if they would have a fresh air mix 1 the right on something like this method is cost you would not be Fuzeka in the same broke as when it came the right the 120 1 its own on the hunt for example right now so it can be read in underslung will create it impossible right now its own other politicians so this is a White House piracy is really a big issue this is why not all the database management system so something like this and win the ball support them than that of some beaches where you should not be OK but the sublime up babymoon should enable of leg and if you are in a world of figures would know that something happens so it's not achieve yoga dusk OK maybe we would also take a break and minute break so 5 by so far great Sunday into
an next public for today is joint Optimization and I'm of cost you very often have to use the results of a petitions wanted petition that with the new means joints and and joined the Renault very expensive some operations so what do we do we optimized drawn up and it really true that depending on the joint planned that to use last the different cost and basically these costs all the old depending on the nature realisation of intermediate results that are part of the joint closest as we know from a relay so that basis to a think I'm include a optimizations we can use different roles for relations other but we know that to join someone and or commuted to so it doesn't matter in what order I'd put the joint tables and they are associated so a rise of Sri rejoin all fully joined it doesn't matter for the result of which joins I'd which which tables light on First but admitted mwave last the influence the intermediate results so we can do is basically weekend evaluate individually each individual joined in any old began rethink of drawings like this a fully joint of joining to and
Buntingford during the 4th time that we could face a bill which he joined 3 where we joined 2 tables 1st into instances and then joined the result together and they come in different shapes and shapes and may really a show different division for 4 months group but even if we consider only joint trees of the show that she was a for example of this is to say that want to do we may still which quality bills and that the medal which it of 1st joined and which is that this is the last to join I'm and that makes a very complex Optimization problem is actually and he ought to sold determining the correct all the best joint reshape with the best assignment off tables is differently to opt mobilisation public and what comes out is that the number of possible joined trees grows exponentially was the number of on relations basically that number he time and fellow of assignment of the different of the different joined tables of the order in which the people of joint of cost of whom the HBO basically faculty many possibilities to do that because the can stop with each player and then joined each other but they were also it will be just basically faculty of and that the set up and 2 are draw Kamaz as basic the free choices consider all possible drawn trees which basically used to be thought about that not of my of basically a possible so
on though drawn algorithms today restrict themselves to a subset of trees and very often does just left the joined so basically joining to tables joining the 1st table drawn 4th table joining the that is basically the them of the of the kind of Tree that consider and then you can use that to the tree shape and fitted with different assignment of tables where you could use your estate to pick a certain shapes of some but if you very interested in parallelization which joint 3 would be a very good choice while what exactly so if you need parallelization basically a bushy choice if you take the Pipelining it's definitely of left the tree draws so I'm not a you want to know more about the drawn on all provides a should look it up in the relation that basis to lecture or some of them are during that fight not anyway and to the point that we have to solve a where houses basically how do we optimized join over stop scheme is not as easy or well said Joe Optimization
never easy but that cause for different algorithms and different terms of efficiency than the mould of the resignation of because of high this last year and look like basically it's a very big effect table that is at the scent of everything that is basically part of every joined and then I have to allow them engine Taibu that connected to the fact and just all some information about your graffiti of some of the products sell if you have an up Aculeate and ask you don't you will have the sale of the sails table at the scent of things and these are the joined attributes that makes all collections of the damage usually do that in the way table in the way across the can also do it in the from clause if you to join me say as don't follow of system time and so on and the 2nd thing that I do in the way of is some selections from so these are basically joined at reviews and the other 1 of selection attributes we just focus on 1 the best places I'm is the author series of has some restrictions or you group rise in the diamond and the and you'll may also put the restrictions it would push
them dial over the joined for this table where they are needed it doesn't help you to make a complex joined between the fact but and geographic location the only interested in in record from Germany then just take those records that off from Germany from the geographic table and joined them was says tables will what the benefit of the so Mbaye basically saying as if you have a civic on G O graphic is Germany it all up sales joined Keogh those exactly the same way as they would to sales joined she the of G you up their father's help you what why would use this want you guessed prop has special unions that the number of results from the start after the joined the intermediate result much smaller than the intermediate result of the strong of and since time in joint Owais has something to do with the number of intermediate results this is said activity of the Joint attributes it's a good idea to push up the restrictions all group buys as far as possible again still I'm even if you do that and then you put everything down to the fluid worked alongside his took an exchange the different joint tables in the right this ways and basically as said the faculty of possibilities to put them together and you could said sales joined Geo joined time to follow the law says on time on the heels of four and so on and into change of just I'm to use it to reduce the actual number of drawn off again he would you the you rely on your sticks and them in a in my world online transaction-processing Systems you see this
usually not a good idea to join tables that I'm not connected by some attributed to of a solo for example FIA geography up joined with time you don't have a single at tributes the care for the joint basically generates took a TV and product which means to build or Lopez of different your graphic patience at different times of the day the big intermediate result of the incident that happened and even if you pushed on the selection but still could be ready Krasselt and a group that has to be of limited by all costs of a new that looking and quality Pete this desperately to looking at start schemes in that allows what happened for such a
ruled out at the world but not as a second thing thing about the rural that he joining tables that she had no dimensions of the channel at review is not a sensible thing to do well the generates into cross followed what happened in in that a well out it sensible and other where houses to join table such on no attributes why not exactly that were basically means joining them in the table which don't carry any information at all in order to other cases we offer with the sales tabled the fact table in the middle in the central of all team up the contains the with different keys to the different demands so the case that tables of be joined when no comment tribute is in place does not happen all to put it differently the fact table should then always be on the leader of the joint which of them will be joined 1st because you need anyway and building all these cut usual part sublime engine and then find driving Investec table does not make sense so again this kind of produce is the number of choices that we have enjoyed for for example if you have a I'm joint offers a sale sector before the geopathic occasions and say we have 10 million records us for the sale stable and in Germany that tend stands out and and generated to sell by double products sold and 20 days and the other to a group has 50 propped up and then by joining the tables was fact 1st we might decide which owned the selected a tea of all the different of tables on sole with 10 straws and Germany is that a significant amount of the for of the says that we have one off we need to selected a team of says table that could suffer established 20 per cent which is to select a witty of also offer before and Germany and a 2 per cent of all sales come from state the all where the rest of the world basic and I'm than the selected to see a small and the intermediate without will still without and 2 million records was stopped was 10 million records with 20 per cent said activity of 20 per cent of the 10 million belong to all the German German sales and doesn't really happy of also soon so of the table but the refuge a but does not see that too many countries it may probably not appeal of good idea to start with and index the doesn't help you either so I'm not you basically do we it is you 1st below the across prop of the by engine table and then you join in the sale stable the fact take while that
help you much better it was set in an old 2 pizza very bad idea to do that because the cross party Devery inefficient could still be the case that it's a good idea and then away houses to do such things right if you dimension tables are very small and if the city selected it between them and the tables this time so called for about the sold and Germany with bulls in Germany polo certainty year for a couple that has of selected as opposed to the millions and millions into entries over fact of this case it could actually be a good idea to do this Cup Tegen product in from the so if I of big time engine tailored very bad idea creating a cut using broader product blows up the intermediate result than joining with the fact able makes it worse stop with the fact they will joined them on the other hand the diamond tables of smell from so the committee's hype all the unique just 1 month although some media but the whole timetable you just this for longer and 1st joining was the fact table for a bad idea for Kent what I would do it would really built those Qupty JM products because the selection the of the selectors and the and joined fact tabled that has the 10 million that entries somebody take the Judum engine with 10 spells at the time de mention this 20 days and approach demand is 20 Prolog so 10 times to a 20 times 50 makes 10 thousand records as an intermediate results yet UK and so we have to underpin and 50 and the and and and holed up in next and self has not to launch an knowledge on between a table of 10 thousand and all states sale of 10 million is a good idea the fluidity of 10 million of care then Guido St turned stole you makes basically from
cricket Germany a 100 million toppled the very bad intermediate result very good intermediate result a pay for this basic 0 the selective reduced much in this case Bandirma this basic you have so
I'm sorry the non sensical of costs is not 100 million but that would mean that we in the Cup he should follow but if we consider that 20 per cent
of the sales were found in Germany will be 20 per cent
to 1 million of the sales tabled by that go for 10 stalls which basically with which to to million he still say mokume and 2 million yen Ten solves and he an intermediate result much more spectacle and on the other hand has already mentioned and the damage across price can become expensive within the
restrictions on the diamond all we have very beat them and take a sell if we would have
thousands doesn't silent customers across product that would be a 100 million records while very good at it but it so well on the rules of the take off with the smallest tables for local with the fact that were 1st all take the fact table as last table but the cut used but it really depends on the on what you actually doing and that is the interesting will go into 1 of the algorithms that applies here for action to help the search joined up for the service said that if we have a smile on mentions that is capable of but the Zamperla and the and the and join this
with the fact that factory will and though it's not the best solution for such as we have if if we have a big name mansion tables and if we have pursued a TV meaning that its stake in the mine and the company will not for sale something going this and they have been Germany improbably 800 pool but thousand of some points if I'm going to do such as the common global head of such as the late in the game the handling of the credit example about even about the safety in the team with such a amount would go he will to 20 thousand has plenty thousand airport is the temple of the book is I'm going to the and saying that the head of Osinniki over 1 thousand about it will be a 20 million records in my Campora tabled the banque to join with my baby effective but this is not the of of a single case tool to solve the close you kind of have found the right to have a big pack there will be a very big day mansions so that solution which has been it has been implemented in the game because of a build up the system in the post here system which is produced the distributed by the end was produced some specialists and he joined the said OK why don't
we introduced a new makes 1 hour affected by on each of the dimensions of the sort we 1st the mix the fact they were do so they will with the tree reduced Übermensch no so I'm going to walk buy a mix it on just the duo 1 index and then I'm going to apply to a specialist and he joined and on the same day Kanepi based on the G or Eyespot have 1 thousand places but 1 said that but are not including the 100 million at the World on an considering its index based on the deal now this is much less than their leading their records if I'm going to fall below this autumn it will be done is to be married in the same region operation defused the Super soon joined which doesn't those who were temporary world but bankers the punters and the idées which figures from the between makes it was sales in the dealers paid East this is busy created and analyst with by these we Damian memory with pointers on the records on sales on silicon disclose operations were this bankers will mean is that they can point to their sales on the high brow which were found in Germany everything can be Estonian in memory because only spur pointers and they beat existed so far so they dance OK came if Web during the election in the wake of the Super joined should still be used as member of table because we might use it was exactly what happens to those we and the cost of buying the and the of as a soul crack would say that we have the jewel in the status cannot which looks something right ahead of Wijewardene the country and interested in the countries
the and the and the records with the country's Germany and the and the index Holmes who on the G why he would the point for the record he Britain just getting the but the same thing but this is what I'm doing is a temporary today but it is a result of a selection and this might celebrities John Reid that the surest way to which have been found in Germany again home and Distributed don't appear in my index fell below for all that he is not interested in interesting because of the selection so this is a lot of stupid joined to pursue a major and would actually spoke now below problems have to do this each of them interested band during to joined with the during the cooked easy in the UK I'd have prefered that the cost because the dimensions of also be so
hewlett is exactly what the described before the I've been doing the same thing open about the fact that it will be next on the public by the again this would be vice because they will be to antique I'm doing the same thing and the and opening the important point about the way after the selection but and this can be big about 1 thousand causes this is my safety and then that was so good might these I'd go the same thing for the time by mentioned I'd get the timetable and the set of ideas and now practically have recertified these entry temporary but a set of ideas will need this was from Germany the over provides I'm interested in and the same goes for the base in the games and the same interested how can idea the to the point was well paid these I'd have to say a camp the money the section of care for this is not the classical reduced Mellor the database in the section this is just the 1st in the section where you want to go to the combined on their ideas from the City and in the 16th and 18th included the jaw graphic against the secret of the babies and by my these from the same stable everything Kaplan's in memory everything these still list of numbers were now it's with 3 with sales own quote spoke wants the point where the River required by the in says the jewel graphic Alegi the point about the way the and the time I'd be I'd want to do so and Reading on the high rise solo possessing using the their memory how to get to the cost them because they are still 1 4 4 4 paper my might be 1 of World by use this these which in the 1st case in the 1st Lucy it can help them I hung on the amount of
demands doodling the section from either the jaw graphical them engine the for about their mentioned for the abandonment these is the 1st case the rise this she says a lot for the scene the 6 and my size of obesity and the fact that not all time to which Juliet Macur by mentioned through for example country where they have got created might then post-election result imposed your Africa both which comprises exactly 1 thousand I'm just going to join list of what they have been memory replaced by the equally to the city as the so hygiene the jewel graphic of hygiene the per about and the Bank of paid whose time the columns it and the band idea anyway doesn't want to see what they were in the bag the and only now I'm going to says the high drive to prefer it is James between this memory but 5 and the 1 thousand popular then the 20 from she would and then there 1 thousand when she went up a is that this is all the model because the would they have been memory has already been the computed with the idea of providing the section so Island and ideas she confirmed with my selection of the same words which have been banned in Germany in the given time for begin game and political section of using this section the growth of slide before now Sozio Germany in memory should take on so says I Germany city by might be about to Jamie interested in Suez by the
Byron interested in whole the intersection exactly the same records I'm interested in the ideas needs to give them needs for this reason I'm just during the and joined them with the dimensions this is right I'm of the trading day of the World Cup the Jan for about and there is still a world to from more big across the differs
during is it the red for 1 of his works what is the place to go the idea it's not that to the world but the it's used as a set in the 1st year of the database of will of the people system sold the solution provided by Jim and not only and it's a great though solutions when actually bump which between a big them mansions and those the but the and create and the fact OK and which them continued to what you do is you just the UK so are the last topic that today will be materialise use and them and the take about materialised use is about that you
select some of the couples which up result of some period which is what we know as a view and every time the truth before you recalculated but results of so high about sporting the results of those of and that is exactly what a materialised you need I'm if you put the Vidima to left you or the you materialise the view in the cash as amazingly fish because you immediately can return results of some 3 and I'm of costs the problem is that the F to maintain the view that if the underlying tables change but as was sent another way out of this is not old to key Systems David don't change very of cell these views can really have long been part of the of the choice of having to be recompute I'm still but what happens every once in a while but is the prose so we learn the new data in the next batch of dead for example where the collective system that controls of material flaw production cycles and every night by making Nietzsche processes and get the dates from about of the overall but that a from from yesterday into my dad a way out this for read Computing most of the views of dating most of the view that what really changed in the view is only that the with suspected yesterday to some views may not be affected and other views containing but also that of its that may be effective in a very specific way so if we could have incremental view maintenance of words that will be a good idea very efficient idea to maintain although he said the Prime materialised few basically used and that a way of closes not true that the same curious opposed to every every 5 seconds to because the most queues and that allows identity precarious that run over long time and a why should your on the same period twice
are could be a different departments are interested in the same thing but that would be wrong or at usually different departments are interested in different aspects of the same kind of that but still they will polls different case all same to pop up of for strategic decisions doesn't need the freshest that at the end all like you might consider some some stop broker application whether that change every every microsecond or something like that is also the view major but for strategic decisions that of yesterday on the back of last week may still be a good thing to do so that rather happens was materialised and that it that of their houses is that you use part of the series as a material you consider the following 2 series are we have to make categories some quantity from product where ideas from fact idea grew by category of here interested in some superb cash for the next 3 might be interested in some of 1st rather than of cost group postal that of so different tables for joined and you interested in different aggregation's uppity on to the streets of Disney you can just compute the results have a mosque directed table and you just put it to get of classic of or so for computer to joins every had agreed with some joined them every spread for with some of the you could use some of the few that gives you pomp of the on site and the and just computer was missing and that is what I'm going to show now that need materialise view that local total sales the takes the property of the deal by the end of the total which is an aggregate of the quantity grew by prototype the and you might be interested interesting thing to not also per property of graphic reason the region from of so if you not all pointed out to various you see that the 1 in only grew by the ball the other only group category all to allow us to do but both so that it is still intact but we need to do for this Foncia through this created skipped products that so we just aggregate all the price of the plot actually is I'm just interested and in the stalls for this from skips told I'm not interested in which the sold and interest in the product those Kruis can be on the committee were you the both need to be some aggregation's but they can use the view so on the issues of the view that we have is the you create a utilization so Rafuse should be materialised what indexes should be computed to date of the weapons then we have to consider the choice of material used if we have a pre what you use can be used to speed up the game and we have the problem of maintenance should be update them to left to every time something changes are the best equipped for not all of the day old until we have an operating all that allows you can be done in an incremental fashioned and so do we have to re compute the whole join or can we can we take them for of a utilization off use of costs have to be a transparent because of the crews are basically interlude rivers and not using the tables that public redefinition but for will be using some of the materials used instead of the full in addition to the speed of and I'm basically would do right at death as it had looked at the execution graph of execution Brassaï exactly what we saw before the operate at which was joined knowledge of the table of the of the elections in it and you have the different joints and then you can have all the tree and them with look about for example the I'm period on the sale of times product that is then joined was geopathic think what could we do well this trio Olivia looks a little
bit like this trio here only that there is a selection and would be the right to do well by don't want to offend the joined Estonia because of this joint is expensive wide under just take this election I put without out them these 2 pubs exactly equivalent and this is but give me a much bigger jointly which results in exactly the same result them before end Ikons use this very efficient he calculated materialised you so I'd just have a single joined left icesave to join still has to do this selection on the on the meticulous OK but still said the join selections as easy accuse indexes 4 said that he joined is very complicated and all of that was made safe Celia time and a basic the idea embezzlers column on a blow to had just take the materialised you as a book and put it in some part of the tree that were needed I'm not has to be done it is it has to be of that replacement so I'm if
I'd change it to a queue to some community cuprite is Isabel replacement if Hugh and queue prime deliver the same result even though of acute acute find used some tourist you and Furyk use based upon this call when the exact that the results should be the same and that the problem of finding those that this is Essar often incubi optimized Asian and the complete as a very hot problem and you conch rearrange free in in or away is on to find out to use a human you have to use some some you sticks and especially for stock to his the some very eristic that I can't say in order to to integrate immature as you into some queued to get this but if replacement is we have to consider a couple of conditions is the view was computed
was amara strictest selection conditioned it doesn't help us because then we still have to go to the base tabled to get the remaining coupled with the use selective of the mature as you has to be brought up and the electricity of opera is a projection of Okri is used it is rather than the materialised use protect so it needs some column that the match you left you does not have called becomes use the who because again would have to go to the based table to get the remaining columns of health of UK and still if we have a protection in Cuba that uses some aggregation function we have to be in a blow to the rise of the same aggregation function also from the material to the material as he was some all of pre aggregated everything to average values for example and I'm interested in and in and out of the some some of the standard deviations 0 1 of anti the origin of that icon not work on aggregate a can of welcome average debt at the end again if have UK selection conditioning queue the selection on condition has to be applied on the M 2 of the of possible this means that if using a meticulous you it should be less restrictive it should have less Projections should not agree gauge to of granularity that is bigger than the cost of the modern media McCreery and it should not have taken out of roles that will be needed in the tree without the full rules that we have with the UK and as a concealed even even if though the Mattila's use not was exactly Curet all part of the 3 Hawkins reuse it still is a selection in M
should be more restrictive than the selection in Kew we could split out during a and the acute you in a and B 1 works on the with 2 left you and the other works on the base tape either we have to see that the elections that we do want to queue a incubi has to be put together and the and and so we have to union the results that we get from those and this was called a mighty book period as opposed to all more mobile query would just put in the relation at the at what the but you that you in as as part of his replacement for some of the drawings the we would just end up with separate offered put them to the FA example if we have got to be some Hugh and we have a materialised few that already pre a cricket sales and profit but we find that the price is kind of a agreed gated he and end
of the restrictions that we put on the sales here is man restrictive than the restrictions that would put on the same scale well just take makes that just makes later but that was made to queries about 1 raucous you which basically is a selection of Hugh and and F M is already in the mid to low you of and askew it as well as wanted by the tree can be very each efficiently agree gated with are at you graphic that Icesave joined the here would have been another joined for the sales and the politics which is replaced by the mature you still stands as ambles more restrictive still need some tough from the basic those that fulfil as F and have already been selected materialist tableside or need them by used the negation of the selection and a used the origin of selection which basically gives me the remaining set up a banque that says this is the amount of couples of well used in the sales and I've some that queue selection that would cut out this set and a half some have and some more restricted selection the cup of this summer a neat Ikon divide Korean to cost 1 that direct the works on assembly sometimes you need all well amid put it differently cost the tourist you could not be a total subset of Hugh but it could be like that selection and a pinch them red left to due to on so that tree I'm interested in that area a pink so why don't the use of this area the last this area is the exact the what's happening in the intersection of F few and M is exactly taking your neck and use all number to rescue because by using the intersection this is what basically is delivered by the meticulous few and APIs arrestees at this part to go so I still have to get this party public added where all these out well that up with St has from the original status a book from the 0 Giles and the part of the rich and says the but fulfil as Hugh but those Donnea a can take out a moat interest and them any more because of audio them or a got them Sywell had not of the selection off them basically this part whose here on the origin of UK at this time there so much to block period is basically the and of both possible eye need the joined with the G table APEC than the union the result of its final like a simple union the result is that the lawbook de might be because wife because this is disjoint a and this is the part makes it this job UK and the rest is that this is a wide and just political that sale of a city duly is indeed where houses but to use it to allow the use of some pretty aggregation's and as we have already seen and we have problems in the building where all the different aggregation's full of is because
it's 2 to the power of and its growing exponentially the their based them and life at the mercy of the cost of this lettuce rolled so you could aggregate everything which is the apex Cuba all you could at the head aggregate only the city only the item only the all city and items city and he is sitting on a demand or you can have a good day basically of the city item in UK and and only 3 diamonds we already have a possibilities to mature enough which would make the band that contains all of the last 7 the was into a book and then a difficult to do that 7 materialised Hughes 1 based table is the based David which have to update or the 7th meticulous use so difficult to do so I'm is I've only 16 them engines a morally at 65 thousand the to use to obviously icon to the game and will be far too much sought after should decide which used to eye that utilize and which you are just computed when needed and I'm with saving and I'm and the answer to that of cost of its at a very difficult problem that depends on the release that I'm interested because of what we should do is to be shut off we should do is we should definitely materialise those possible usury that we can afford to recompute things when they are not very of computer and to make things worse off the different dimensions classification of to we could have paid up date Mom you the aggregation and also Luis could be pre
computer of the different dimensions of different granularity in the dam
and places where the products between different dimension make a much higher number of consider for example if we have secured lettuce for just 2 diamonds and was a product category for a family of 4 Rubin not to go which makes for a different levels of granulartiy and we have to do a graphic locations would maybe 5 levels than we were already on the obtained more than 500 notes that would be different possible but he left to use of can't his just to dimensions with for the with for defence of quest to weekend some kind of reduced that because some of the giants of the aggregation can be used in different Christopher example if we have joined the considers optical Grubin category we can't you meet optical from just by grouping the group category that we have optical group we can immediately get the optical just by agreed getting all the group inflammation while the optical category just by migrating the categories of of some of these have functionally tendencies to some of us and of course we don't need to stall over from but just a couple from the very often used the but it looks like that basically if we have different countries and different products and only 3 of them we are ready we all of at the order of components so aggregating everything which makes a full plus 1 times 5 plus point 30 different possibilities Stallman to and now we have to
pick which 1 of the good points of for example the 1 may be very good and this 1 may be very good that if I'd stalled this fund than or also storing this want put probably be a bad idea because exactly Ikon immediately arrive at from this materialised used by a great gating over the product category of the piece he thought it depends what you want to do and would this makes a problem you Hobday as it is in choosing the PM at the view was to be materialised means we all with limited cash over the all with limited main memory so we come we desperately conference that you Olazabal we should think about what brings the best benefit for all of other basically 2 ways of doing that but 1 of the state choice about way just say OK the database administrators to stay at some point out which you should be sold and which for you consider the crew load of some point to take the measure of the last
half used and recompute what would have been good choices to materialise and the and that that was made to and the set up of the 2 left to 32 was just remains unmollified until next fresh sold off the half year period and look at the cruel all at 1 of is and and you choose the new set off of joint that should be made to I'm while the problems that immediately comes into into into play here is that the set of chosen to the fields and with the 1st to a time intervals because you do it once every month old ones every half year old 1 of but your and consider this interval for for the choice if not a very Dynamic Crosets so obviously the other way to do it is a dynamic of choice you just Toulouse over time it is not at all what I going on and want to find that some the tourist the Piedmont beneficial and some view that you currently old during the just exchanges while the spread of a new 1 and skipped the old 1 more dynamic of costs what can happen here is that if you have a really patents combined for with refreshed you materialised use and never get sick intent on them because the old crews are running again difficult to see what the idea would have less opposite static choice so you choose what would used to materialise by by looking at the benefits that you expect this Matua future in hominy cuties the you half in the dispute he is a difficult to compute of or canopy derived immediately from from some based a but the cheaply or very easy I'm and for determining that you need to cost function but the cost to the right this you to compute the you for what does every year I save by using the your to smell savings probably not a good candlelit for the to they should be C savings from abroad and the woman who was on the other hand half used so often do you benefit from the savings great savings that all only part of the single Phelia's but once a year to the system following not a very good choice to the tree on the identity of many who are working on a mature feud that doesn't safe to much it might it up from the multiplication from so that cost function also request for pre safe by using the materialist you and what would it cost to recompute amid the view from the frequency of the Tree hominy curious to benefit from this view and the maintenance costs of the view that it is too expensive to keep this you update developed from basic ingredients and as we know from algorithms and best that this is a classic knapsack of West the knapsack was the match that he
as exactly that you may memory all your cash and that is all it is a restricted so that sex and of what is what you put in your knapsack but obviously the space you need for which he realised the thought the joined up and follow the view to from I'm so we have a maximum standards by main memory cash and the costs that would that it would need to store each note of the left of different a condition that we can do and I'm of cost knapsack problems unknown to be and the complete so we count have correct solution about what he usually do is used read strategy for you include the last of cuboids so what can be backed at expected ability of multi the maximum but size and then you stop with the most efficient choices put that into the knapsack which would uses the next excise of meaning signs and then you take the next best candid at off the betting that the Fed and the into the maps that under the SEC full as a basic at the start of a half point to if you have some of materialization off 3 different at should be you have to buy stake which space for materialise you off to celebrate for for this is that the is and possible
aggregation so that 12 mind 50 regions and their 100 groups that is easy to compute if you spread everything 12 time 50 times 200 might give you 6 of if you start for only the mind of a 12th entries only the regions of the occasions of regions 50 50 record agreed cases of groups 200 if you use a aggregation of 50 regions by 12 months would also already result in 600 if you choose to a good everything would give you success of what goes and the and the SEC have to sell and endeavour was stops and but this week we have the cost functions that shows how many crews posted hominy of you want but the and home Manea high much by by average did series benefit from using the mature to end of the longest into the fund so for example you might know of that the seed had and aggregation off a 1 day to find with 600 records of good idea of sold 40 take is to take this mature you walk and became a key later
from an off the is the again calculated directly the month from left by a group dating of the region you can agree that you can Calculate directly to the region from the by a group making the monk full you could use it to aggravate month and regions and then compute the apex Cuban up and so that the next step but it's not a very good idea to put either a 1 8 2 or into the knapsack doesn't help you so what would you choose for example you could choose the street because the tree is on ice we have very high frequency another from the crew of again that comedy do you have your 600 here have 100 Shia still remaining 13 under what might fit in the region has might fit in but it's dependent the can can Calculate something from the 5 thousand you don't fit and the 6 thousand he adopted an eye he is still some way off a 500 that we can materialise so we can't use that which makes it basically I'm a 100 600 and 12 100 is exactly 2 thousand within on at the so that no more than it hundreds of members of the 100 of the last went right thought we didn't feel on knapsack yet but with the remaining 100 Weekend The the get the five fold in of 6 thousand and and that the of those 2 wicket that if we wanted to and from the UK and they sitting on the deck of a function the interesting part is not the algorithm obviously the interesting time is in the cost of each if you would cost function you decide for very good materialization to the cost from you might end up with some of this was suboptimal mature so well that the problem was that choice because of the automatic let applications are very often intact if up so you around a set of criticism of the way
and there is some sort difficult because of the fuel ways comes to the same base that in different aspects of this looks more like cashing and like materialization support for good idea cache of you rather than the few that cash last year with a lot of fun to mature accompanied few and some even if you would be able to go to to learn the group at an off some time you don't know whether this Croesus will go on so will this decision support problem belonged to a problem but it's off a week of the month over just a one off the best want to know how all sales and mould America all some Hall proceedings all of which closed on the ball from recompress so I'm it sits at the edge kind of interesting to see how a man things will remain used but we have no idea about I'm very often also modification over that leads to a higher up the road efforts so pudding really filling the maps of this not good idea because we still doing some views that might not be too useful on but you have to be updated if something not for use should be I'm still also thought Applications there may be some truth about 30 operated under the of on and those who use the use their should definitely be free computed and bomb brings us to the 2nd possibility the dynamic choice of the used when the basic you do with your money it all accused the executed over time and you maintain a mature few professing basically means which few has been used for and whole recently then you
look at the plans and in the background from a algorithm that exploded the inflammation now come from aplomb actively decide from the view that have to be materialised reorganised the existing views and then you in a good state and you can look at the load again and see if something changes so whenever you could land the York Cremona talk that we find the view that would in the light of the recent developments and that a Wells result in that if you choose to go on its old change that basic to what happened at the same principle as cashing so you learn something into the cash and at some point you have to look to remove it from the cash it is somehow evicted by some of the transaction so what would it be the 1st and a you will know some of the very very frequently used
strategies for cash the victim Silvio the of these recently used or of the of the most often you'll strategies all something St this that it was happening here you Calculate the benefit in time after last success that is the reason used basically in the frequency Wolfman most often you see the size of the meticulous you so admitted that he was smell it may take up less space and that which is good of cost of new calculation of actualisation will produce from to the number of grid set up on was you also the some of the to you use the number of trees that could be answer to look at possible furious that he could expect of that these off put together and build basic your cost from than you Dynamic the update the cash although but of the 2 as you space and in each step the benefit of each were to you in the cash as well as the view that is created by the are calculated and if you find that
some fury creates a view that is a beneficial than something in your cash if the those things from the cash put in what you learn from the the results of the a basic as material that is how it works St and keep your but you have to face up to buy the benefit of the of the crew said you have and whenever you prefer the new period decide if something should be victory or if you find was what you have the cash at the time of dusty that the UK and the maintenance of a two left you of the final topic for today's so about how do you keep the materialise you up to date with the based tables Hollywood to fresh when the underlying tables with fresh with some new data comes a and Howell to refurbish of you in response to a change in the light this the and linked some updates of every off their wealth but and hold new to refreshed view a piece from a shaving time dubbed the Dome notes that for the good of idea to go now because they high that we go so you can either although member ways to of update and maintain materialise you is that you recomputed on every update in the basic of is the those cattle range solution because of the five recompute 5 law
updates on Monday stables the No recomputed joint every 5 minutes which had cancels the benefits that a by using the joint because of every crew that use some them is kind of a recalculation off the join them a code of recalculated from stop doesn't the help you but even can can deteriorate for 4 months if you are update Load as of large Roman your current load as much more fish to re computer joint went creating than when updated what it out with updating the view 1 of the reasons of 50 for will to send them out as the South when your update Load is actually higher than you really load the new and the updating all the view that are not needed by care is to spend a lot of time for nothing has helped to promote and that case it's a very bad solution to Lewis is definitely not the best solution I'm backdrop is the is the incremental view update sold only part of based stable changes and only those few that use these polyps should be updated to have to be paid when the basic the do is now and you look at the changes in the database relations and use that to compute which used to be updated and the changes to be done from a mature as
you is recomputed into some the 2 as you the Prime as basically you take out everything that has changed and the origin Achiltibuie and to add what has been inserted into the origin of air use you may contain aggregates this sometimes not easy and you haven't for example you you have the average price of coffee of products now you take up some product was the average change difficult sometimes you can use them to get up to that has happened but if you use just inebriation on set of couples when you can do that because they say if you want to say so I'm you can choose to do it immediately when a chance action modified see on Taylor to of the underlying Tevez this modified you look which use are effective and community modified of goods Reviews allways
consistent the freshest I'm that he operates take a long time because I qualification to be changed to give with a 2nd possibility would be to do with the 1st some time later when the separate infections so you stop was every change a transaction and every I'll or something like that you just 3 fresh to give you full when you have some Computing Pul Singh slow down on the little to thought we Computing from OM advantage its case is very well because you can maintain the view was sold as he of computing Paul suspect Computing polo 30th she and but you can still do you curious without updating the views on the world view so if the mistake was not to be a me now would do it for stockbroking but for most of the decisions that have and that a way doesn't really medal of views of the and Iolo draw has missed the last 3 updates of disadvantaged your income system that if it employs that you that as fresh stockbroking or something the and the stiffness of public of the seed the before but for fresh comes of free flavour of you can do a lazy deferred for fresh so acute delayed the refresh on till the view was actually period which means that its part that part of the update pro sense but part of the with their choice when Chris off left frequent and updates this of a good choice but update are more frequent than could you could eukaryotic lead every or refresh my views very good idea if you are not interested in that of the 1st to consistent and that is what most that a Wells action to because the useful strategic decisions rather than 1 line broking of so this is what you should do all you could do even based so after fixed number of updates offers something offer the best close to be really all the event may be began to find the list of events an on these events that he was well and that it fell to date is that no questions left for a run knows about petitioning so everyone knows about joined
everybody knows about mature for use or you need to know to handle with the warehouse in your company by next time moved to be concerned with the way solo go directly into O'Laughlin's see some of the slicing and facing going up the interesting to see who will talk about the scale for that houses and the multi-dimensional the Languages like and the UK's which will be the most popular 1 thankfully and took and not it can save
Retrievalsprache
Mereologie
Leistungsbewertung
Datensichtgerät
Fakultät <Mathematik>
Gewichtete Summe
Schreiben <Datenverarbeitung>
Information
Bildschirmfenster
Rechenbuch
Formale Semantik
Freeware
Hausdorff-Dimension
Algorithmus
Vorzeichen <Mathematik>
Gruppe <Mathematik>
Konditionszahl
Computersicherheit
Statistische Analyse
Kompakter Raum
Eichtheorie
Kontrollstruktur
Ordnung <Mathematik>
Gerade
Phasenumwandlung
Auswahlaxiom
Caching
Prinzip der gleichmäßigen Beschränktheit
Distributionstheorie
Automatische Indexierung
Vervollständigung <Mathematik>
Datennetz
Computersicherheit
Optimierungsproblem
Biprodukt
Entscheidungstheorie
Quader
Tupel
Einheit <Mathematik>
Würfel
Trennschärfe <Statistik>
Ordnung <Mathematik>
Tabelle <Informatik>
Lesen <Datenverarbeitung>
Subtraktion
Multiplikation
Selbst organisierendes System
Kurvenanpassung
Datenhaltung
Demoszene <Programmierung>
Gittermodell
Variable
Spieltheorie
Fächer <Mathematik>
Endogene Variable
Retrievalsprache
Vererbungshierarchie
Spezifisches Volumen
Tabelle <Informatik>
Soundverarbeitung
Eindringerkennung
Videospiel
Konvexe Hülle
Konkordanz <Mathematik>
Vererbungshierarchie
Verzweigendes Programm
Softwarewerkzeug
Gasströmung
Binder <Informatik>
EINKAUF <Programm>
Hill-Differentialgleichung
Datensatz
Netzwerktopologie
Offene Menge
Caching
Wiederherstellung <Informatik>
Wort <Informatik>
Faktor <Algebra>
Heuristik
Prozess <Physik>
Klassische Physik
Element <Mathematik>
Natürliche Zahl
Familie <Mathematik>
Gruppenkeim
Versionsverwaltung
Element <Mathematik>
Computerunterstütztes Verfahren
Technische Optik
Komplex <Algebra>
Inzidenzalgebra
Analysis
Eins
Metropolitan area network
Maßstab
Gruppentheorie
Total <Mathematik>
Uniforme Struktur
Bildschirmfenster
MIDI <Musikelektronik>
Flächeninhalt
Steifes Anfangswertproblem
Wurzel <Mathematik>
Parallele Schnittstelle
Schnitt <Graphentheorie>
Umwandlungsenthalpie
Lineares Funktional
Schießverfahren
Physikalischer Effekt
Datenhaltung
Reihe
Speicher <Informatik>
Systemaufruf
Übergang
Nummerung
Plot <Graphische Darstellung>
Frequenz
Widerspruchsfreiheit
Variable
Algorithmische Programmiersprache
Spannweite <Stochastik>
Arithmetisches Mittel
Teilmenge
Transaktionsverwaltung
Druckverlauf
Verknüpfungsglied
Datenstruktur
Projektive Ebene
Garbentheorie
Versionsverwaltung
Schlüsselverwaltung
Standardabweichung
Aggregatzustand
Web Site
Total <Mathematik>
Systemverwaltung
Wellenlehre
Fluid
Schaltnetz
Physikalismus
Gruppenoperation
Rucksackproblem
Zellularer Automat
Nichtlinearer Operator
Term
Code
Division
Wiederherstellung <Informatik>
Data Mining
Physikalisches System
Mittelwert
Notebook-Computer
Arbeitsplatzcomputer
Stichprobenumfang
Zählen
CMM <Software Engineering>
Biprodukt
Strom <Mathematik>
Operations Research
Ereignishorizont
Basisvektor
Hardware
Schreib-Lese-Kopf
SISP
Fundamentalsatz der Algebra
Fehlermeldung
Wald <Graphentheorie>
Approximationstheorie
Hardy-Raum
Verschränktes Produkt
Relativitätstheorie
Primideal
Physikalisches System
Lineares Funktional
Umsetzung <Informatik>
Sichtenkonzept
Gerade
Quick-Sort
Office-Paket
Flächeninhalt
Last
Hypermedia
Dreiecksfreier Graph
Räumliche Anordnung
Mini-Disc
Greedy-Algorithmus
Hydrostatik
Einfügungsdämpfung
Momentenproblem
Extrempunkt
Desintegration <Mathematik>
Definite-Clause-Grammar
Nebenbedingung
Datenmanagement
Orakel <Informatik>
Extrempunkt
Gesetz <Physik>
Raum-Zeit
Computeranimation
Homepage
Netzwerktopologie
Negative Zahl
Standardabweichung
Softwarewartung
Typentheorie
Trennschärfe <Statistik>
Kommutativgesetz
Information Retrieval
Vorlesung/Konferenz
Materialisation <Physik>
Metropolitan area network
Softwaretest
Addition
Shape <Informatik>
Oval
Sichtenkonzept
Kategorie <Mathematik>
Magnetbandlaufwerk
Gebäude <Mathematik>
Güte der Anpassung
Strömungsrichtung
Zeiger <Informatik>
Instantiierung
Rechnen
Ereignishorizont
Reihe
Rhombus <Mathematik>
Dienst <Informatik>
Wechselsprung
Menge
Rechter Winkel
Zahlenbereich
Konditionszahl
Festspeicher
Ein-Ausgabe
Pinching
Heegaard-Zerlegung
Server
Entscheidungsunterstützungssystem
Instantiierung
Familie <Mathematik>
Relationale Datenbank
Partitionsfunktion
Maschinenschreiben
Stabilitätstheorie <Logik>
Teilmenge
Mathematische Logik
Hash-Algorithmus
Mathematisierung
Rechteck
Abgeschlossene Menge
Parallelverarbeitung
Whiteboard
Überlagerung <Mathematik>
Mailing-Liste
Domain-Name
Informationsmodellierung
Spannweite <Stochastik>
Globale Optimierung
Datentyp
Hash-Algorithmus
Verband <Mathematik>
Datenstruktur
Maßerweiterung
Parallele Schnittstelle
Attributierte Grammatik
Algorithmus
Solitärspiel
Zehn
Materialisation <Physik>
Raum-Zeit
Axonometrie
Default
Schlussregel
Partitionsfunktion
Menge
Chipkarte
MIDI <Musikelektronik>
Auswahlaxiom
Dimensionsanalyse
Selbstrepräsentation
Attributierte Grammatik
Data Mining
Resultante
Bit
Konfiguration <Informatik>
Punkt
Adressraum
Kartesische Koordinaten
Computer
Übergang
Videokonferenz
Einheit <Mathematik>
Datenmanagement
Prozess <Informatik>
Mixed Reality
Figurierte Zahl
Public-domain-Software
Einflussgröße
Prototyping
Serviceorientierte Architektur
Folge <Mathematik>
Zentrische Streckung
Nichtlinearer Operator
Prozess <Informatik>
Abelsche Kategorie
Element <Gruppentheorie>
Stellenring
Klassische Physik
Globale Optimierung
Gleichheitszeichen
Knotenmenge
Dialekt
Teilbarkeit
Gesetz <Physik>
Rechenschieber
Softwarewartung
Mustersprache
Vertikale
Gruppenkeim
Funktion <Mathematik>
Twitter <Softwareplattform>
Verschlingung
Geschlecht <Mathematik>
Phasenumwandlung
Strategisches Spiel
Information
URL
p-Block
Portscanner
Quader
Stab
Hausdorff-Dimension
Taupunkt
Zahlenbereich
Kolmogorov-Komplexität
E-Mail
Heegaard-Zerlegung
Multi-Tier-Architektur
Graph
Benutzerbeteiligung
Datensatz
Multiplikation
Adressraum
Kommunalität
Zusammenhängender Graph
Warteschlange
Ordnungsreduktion
Softwareentwickler
Zeiger <Informatik>
Transaktionsverwaltung
Hilfesystem
Implementierung
Leistungsbewertung
Autorisierung
Prädikat <Logik>
Relationale Datenbank
Matching <Graphentheorie>
Graph
Diskretes System
Softwarepiraterie
Zwei
Geschlecht <Mathematik>
Systemverwaltung
Einfache Genauigkeit
Mailing-Liste
Migration <Informatik>
Objektklasse
Frequenz
Hochdruck
Endogene Variable
Mapping <Computergraphik>
Assoziativgesetz
Mereologie
Basisvektor
Verbandstheorie
Walsh-Funktionensystem
Binäre Relation
Stapelverarbeitung
Shape <Informatik>

Metadaten

Formale Metadaten

Titel Optimization (25.11.2010)
Serientitel Data Warehousing and Data Mining Techniques (WS 2010/2011)
Teil 5
Anzahl der Teile 13
Autor Balke, Wolf-Tilo
Mitwirkende Homoceanu, Silviu
Lizenz CC-Namensnennung - keine kommerzielle Nutzung 3.0 Deutschland:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen und nicht-kommerziellen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/332
Herausgeber Technische Universität Braunschweig, Institut für Informationssysteme
Erscheinungsjahr 2010
Sprache Englisch
Produzent Technische Universität Braunschweig
Institut für Informationssysteme
Balke, Wolf-Tilo
Produktionsjahr 2011
Produktionsort Braunschweig

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract In this course, we examine the aspects regarding building maintaining and operating data warehouses as well as give an insight to the main knowledge discovery techniques. The course deals with basic issues like storage of the data, execution of the analytical queries and data mining procedures. Course will be tought completly in English. The general structure of the course is: Typical dw use case scenarios Basic architecture of dw Data modelling on a conceptual, logical and physical level Multidimensional E/R modelling Cubes, dimensions, measures Query processing, OLAP queries (OLAP vs OLTP), roll-up, drill down, slice, dice, pivot MOLAP, ROLAP, HOLAP SQL99 OLAP operators, MDX Snowflake, star and starflake schemas for relational storage Multimedia physical storage (linearization) DW Indexing as search optimization mean: R-Trees, UB-Trees, Bitmap indexes Other optimization procedures: data partitioning, star join optimization, materialized views ETL Association rule mining, sequence patterns, time series Classification: Decision trees, naive Bayes classifications, SVM Cluster analysis: K-means, hierarchical clustering, aglomerative clustering, outlier analysis

Ähnliche Filme

Loading...
Feedback