Bestand wählen
Merken

Multi-Tenancy in PostgreSQL

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
for for this Australia have been working out with posters for the past 10 years on and off and did some development mostly due devious all affected the and this is my 2nd PD consul volatility you competitive to that many people and so what we are going to look at today's some you know how we can achieve multi-tenant samples because you know what's available today what can be done and what more needs to be done for different use cases that may be out there and part tutorial 5 just tossing off ideas and because we had begun conference there's still a still an unconference mode so please feel free to jump in anytime
and so looking at why do we need multi tendency and I'm looking at it from with its point of the obviously and for this is a big hard red winter and rebuilt really big machines and so hardware consolidation is something that really interest us and our customers and you may think that you know these days so not many people are buying bigger than bigger hardware that more Jewish realization cloud and things like that that is true to some extent but there are still the emissions being built up alliances being built we do have a poster plants and the still using it what do they like it would be stopped lentils obviously it's the plug-and-play is the simplicity it's the productivity gains that come with that and so hardware consolidation is a bit case so when you buy a big emission you want to put more and more an more instance of injury hot water when you are running on if you want to put more and more and you want all of them to play nicely with so you don't want them 1 killing each other they all said they're sharing their resources and things like that and so that that is 1 of the big attractions for us when we look at multi-tenancy and resource-sharing comes out of that all of those instances hasta share on the resources and then cloud solutions Cloud solutions is very them look at it in a couple of font slides off of this and what what that means is that you know are there are solutions like obvious out there are obvious being very public but there of a number of other data resources so we solutions in many private clubs which have not sophisticated obvious but this still needs something at the database level to make it really efficient so that's where the need for multi multi-tenancy constant so
when someone is looking for multi considered what there really and look for and obviously securities security not in the traditional sense they want bit isolation that's that's 1 of the key points not just that isolation that data resource isolation of small so that users they have to be separated 1 use of shouldn't know there is someone else and by this name in the same instance so you be has to be clearly isolated so my use of my data the thing is happening and then comes optimal resource utilization everyone wants to build a system that is being used optimally but they don't want to read some of it has to be perfectly tuned out perfectly and you are perfectly evaluating the service so that's what the optimal resource utilization you buy a big system of small system doesn't matter and you want these services finding inside it to really useful but it's also smart to efficiently without having to you know compete with each other this is an ideal world obviously and then application transparency is no no big deal it already done you cannot really do this outside with the fall of the people who live in the schools won't be exactly it I do have that in their later yeah so there is some of the resource allocation as well as the resource restriction here of what what I found the points of the and then there is our obligation transparency it's no big deal but you can already do that applications with posters QL it's really good at it is this kind of connecting to the maybe the yeah and what is there but so the application doesn't know that it is connecting to it be passed with thousands of databases and it's all it's this connecting into a single instance with single that others for that already happened but nevertheless that's 1 of the things that's needed so this
is looking at the use case it's a bit lower between documents and use cases and but nevertheless what other use cases why do people come apart from the points we looked at in the fast like what other things too little look for why do they want to go into this area uniform platforms by platforms across cloud and on-premises if you look at the young articles database appliances now they're actually held OpenStack inside it's exactly the same platform as they haven't clock if you go to Oracle Cloud is built on OpenStack the services that they don't lose sight the best services are allowable in there called platform is exactly as the 1 you get in the it number will more small differences there but mostly it is the same platform why do they do that the customer they want to have a uniform experience they don't want to have any of you who is with a specialized in the cloud for provisioning so you know managing cloud instances but then have need to have a completely different skills said to manage on-premise instances so they 1 so that's the and that should be appliances will this moving as well and the appliances read to have with posters Q is also looking at that space so everything has to be uniform you can move across cloud and on-premises with having to take into any of your practice do most it seamlessly back and forth then there is the other other activities generic infrastructure for BBC again you don't want to have speech if you don't have specialized hardware you don't have specialist up alliances of platform you still want to do many of the the things you would do in the cloud with Monday 10 in smaller systems so well and and then easier integration with other business and IT sector doesn't fit with multi-tenancy but nevertheless you want your systems to play with and you look back up and storage infrastructure for example of indication infrastructure for example so they all have to work together because most of those things that are already in a multi-tenant at rate can be restricted based on sort and instances and things like that so what do we
mean by multi-tenant depending upon who you talk to you get different interpretations and they all can mean a really them really different things the basic 1 is obviously have a single piece of hardware you want to run multiple instances in this is what we do already imposed skill obviously that's that's the purest form of but then if you look at from the candidate of his point of view of the new 1 have 1 instance but you don't want to have so what else self-contained databases you want to have a database with the users within the database they don't see the harm in the catalogs of a candidate of is that not shared among each other so all those things it's completely self-contained and user logging into a database has no idea what other databases exist in the system this this this like element of this definition is somewhat more falsely people considered us multitone and what we have and the faster and that's a single piece of hardware of multiple and multiple instances is actually a lot about to achieve the same thing but and then you want to have it fully multi-tenant 1 this is what you will fall then there is the extremely you Nietzsche and recommend it is 1 instance multiple customers multiple instances and it's something like nativist or call for example and it's 1 and an instance and with 1 database as far as you can view with so table spread across multiple customers multiple years of and technical in this same same base using global security following usability and all those things that this is not going to happen for 99 per cent of that uses this document is not going to exist but assisted many people looking at time the building private obvious solutions dislike him may become more and more common as well
OK so let's look at the single piece of hardware and multiple instances and this is what you can already do but even if you can even if it can only be done there is really no way to control how much resources they consume and how they interact with each other for example you know I do have that in dislikes later I can hack it did so what I'm talking here is from a database point of from within posters cable and what you can do you want to be able to say especially in a cloud environment when you want to set aside you understand if I allocate to CPU sort of relative this much memory to this instance is always going to have access to that much 1 of them there is not going to be hoarding by another instance here and then you also want to restrict enough so you have the minimum level and then you also have a maximum inner ucla's find a seller you define minimum this is the set of resources you're going to have but maximum is also there so bad that needs to be set as well and then resource sharing and they have and they don't go along well with each other but you can have a high degree of situations where if there are free resources and the system why not allocated to others no 1 is using it and if someone else wants to have that show allocated to them so there are 3 things and when we are talking about the resources and what resources fairly common CPU memory and disk you want to limit and restrict the usage of and CPU and memory I'm fairly and self explanatory and this is the the stories and the tenant as well so this is just
some very simplistic view of how the mind turns would look like and and this is 1 of the 2 most in line obviously they have premiums on a hardware and is running some sort of from anywhere on top of it and if you look at the graph shows the All-Ireland perfectly they have different the periods but I am not at the same time obviously and so you have a large amount of resources needed CPU or memory of what sitting idle at different points in time so if you are going to pull them together in a bigger than 1 and then why not and put it in a multi-tenant environment now why you want to prove but apart from the obvious choices of resource allocation of the new states why do you want to combine it in all in 1 region there so recent and beyond swapping medium and putting and and follow up all these things which are available in many of the and what so platforms that's relevant the states you can make use of that it becomes much more simpler having for 1 single medium compared to having to pull 3 5 7 years shorter depending upon the worst platform you to those things can be easy and but from a database point of view and the same but it can be the mentalities and other sports that have modeling really get it yet so this is this is the multi tendency and this case so you have the resources they need to be available all the time so what
what would you what to read you haven't posters skill already is that we do have survived instances that we can already do strong ethnic isolation once you connect to a particular instance you have no ability of 40 so relevant the evidence fairly straightforward can already do that then we also have separate databases again from a data point of view once you connect to a database yet you don't see other databases but much fully can see but you can cross-link and and it's simplify data management and then you can also use this to that of multiple users but you still want to connect to connect them into a single data is you can split them into separate schemas so that is also already available and but it does have obviously a lot to me and and so and especially when we don't have automatic use creation for of skin and saying does come with its own have but nevertheless it is that you can easily write scripts scripts are already available to those things but all of the above 0 share the hardware resources this to compete with each other so the solutions we have so what can
you do see groups so and I find out how many people have used this associated groups is a Linux constructed so what you can do is you can pool resources and a multiple of the it landed there is a list of things which you can work based on CPU based on memory and the I will through for all this can be grouped together and then you can look at them to a process so as simple as that and you do and it is not like and you can only allocated to a pre-learning process that you can actually define a process that will start in the future and then say that process when it starts should go into this group so that some you said the limits and it's it's fairly straightforward it some if you look at the configuration file it's very similar to speed up on that it's that's tight anyway and so this can already be done and and pretty sure many people use and you of running multiple instances in a big big system and but it does have its limitations as well as I said it's purely were mainly process states but if you look at it from a database point of view wouldn't it be better if you could limited from a user point based on user's databases you understand this particle emitter misuse of when he comics starts this particular process so that connection Klein connections belong to this use of the system to be should go into so not understandable SKOS processes it's just afterglow crosses that is connecting to a particle emitter with a particularly as well as the 1 in this whole things like I lost all share in the restricted 1 process in the way that we can use this nice to we help people all of all these years here or 1 that will work of use of I would use it to get all full of these probabilities in this instance the probability of all that there is a set of users he's on the use of limited down slightly by the user really pulls the process yet Banyan yet that's that's I mean something interesting this yet so the user can access to the database obviously so then you have that connection then you have that translates into up fossils on small process or yet spots the all the all eligible people that's the yeah that's that's yeah yeah this is all part of the world where where you know you you you you you you you you you you you you know that that was the problem is is that 1 that 1 1 of the things that happened was that a the like you would you would like you us to work with here that the universe go never end up being able to use the data that was held at the heart of what it is yeah that that's that's definitely the new point as well you know because as much as you can do it easily and how it works with outside of the database you still want to have something and that translates within the users and the database of you know the posters convinced so the part of the hi would you know I would say yes but not in a serious way you know I would say that we have been playing around with that we don't have anything very of solution the ladies recently yet I I have to say really set up separate begins by by giving them priority so works a lot better than you you have won the and it has a low priority but as long as nobody else need at the beginning we also shown that the use of the new book that had not there like bearing looks like a whole new way of doing it all comes from you know yeah yeah so yeah so the city groups of
basically kernels and what I was talking about still wouldn't it be better if we can control it or have some sort of link with the database objects itself so that we can control it better so
when we come back to that multi-tenant use case to other databases and I think we looked at this 1 cannot find all the other and not quite involves goes to user can access the database data but in multi we want to keep them completely separate
so 1 option is to have I mean this has been an attentive at the end so a few discussions and the hackers group but nevertheless I'll bring it up again and yet so ideally we want to have a bit of this a specific uses roles such that use that have limited to just within the letter days so 1 day you could do that is to create a role to to have and the young and scope of that that'll so it could be classed avoided could be that of his wife and so so for explanatory so possible where users can create any database in the past and database and specific users can only connected to that database and the rules associated with was cannot see the data not belonging to the instance goes without saying but also always and again the course always has been how we treat the shared catalogs and how cool gets to see what from that shit catalogs and how we split across the databases so that the data that is not specific to that database is not permissible and so
maybe of securities the answer we do have of security and 9 . 5 now but and it not available on system tables yet but maybe maximal evolutionary history extend that and into into this particle use case we want to I want you to think of better when you're talking about the sky was and what role in some of the system doesn't do through the more you know that we're in this together what we actually get right so you can still try grant you know explain this lack whatever to some user you can detect that is that this this is an issue with the rest of the question is do you but you know where and that becomes an apple down the digital world when you when you want to all the has the same username maybe has somebody else that's right that is going useful for we really need to have a solution of the galley part of the University of his so this will be the other objects in the you know the the difference here is going to have a case where the you have only 1 person or a small group of readings in in your in your view on spot for a reading of schemas and so that you can get away with this is the 1 that you worry about what was going on in the in the year the series see here and now stands at the family setting to decide that the reason that I think because it is right there is allowed to get the best of our knowledge of it all in there if you don't want it means you will probably include except when all around the wall and its child is just all of the problem but if you don't have PTX-loaded you due to get the data that is directly related to the then that we should work as well but it won't be of use to them where you want to call it the trees in this culture this is of our problem I actually do have a slight tools let's let alone that promise the that the 2nd is a don't have a solution I just do have a slight like this here on the and yet so low security is something really looking and also how that can be extended across the system catalogs to really limit that information and so 1 of the options we were looking at to select using object is less than the 1st normal object are based on the use of ideas so that dog looks something like this
concept of please had so this not a solution it and so the idea is to just create a policy during the the debates on it and it's it's not as I said it's it's not completely just at the area and but this is 1 thing 1 possibility that we could do so because it is done and they needed time than users who don't want to have the those limitations doesn't have that'll work but there's still get to see everything else and so this is based on the the Wallace nominal to and when the user was created and all those things and so this is a possibility and in this connection probably be done all over the you not of the ones of policies the the yeah it doesn't it it doesn't scale I agree and but you get the ability of you that question do we care about you know the name really most of the that is the reference in the in the freedom of the application that that that that we may soon be able to then you can see the variation in the quality of the environment along yeah it's it's more like other use of trying to send a problem and it depends on what you return back you know it at the end of the year where you have like the overlapping problem you know that classes are just some of the letters true there that's the reason this ship roles databases metallurgists that names namespace none namespaces not shared and to share of the global but the shares period 1 the of the 1 that is those that was 1 of the 1st ones like the 1 in the middle and that's where the problem arising from the Treaty that that is the policy the only see during the period of 1 you won't work with you and you go right on the 1 of the things that are themselves and the people the world of the that that's what you get there and there was a lot of no violence this year and this knowledge is used in a lot of their race you after you going to do that the the more you it is greater than that that users to can be used to it yeah that is said to be the next level grant would be the next that that the watchers of mortgages on their the and a lot of people and and not not from across the world that was used in this in this and there but reversible becomes tricky is when when you will have and when you want to create a so by use of of the database itself because now you have a super user for the class of and then on the simple use of of data is what sort of flights to be the best you can see the there is an and so that that but that comes as well and it's easy to say and we can limit that database so by to exactly the same set of operators of also so per user has but contained within that its meaning they can create other users they can create others so when you 1st it but it's still might to walk and but yet once once we have this constant and I think that that is easily solvable that would be the least it's just setting the policy that says it all you and you want and things like that the in 1st part of the reason that word in there is only a very close to this line it so that the figure of way that only 1 reason for it was you know what you call it wasn't 1 of you will always have a fair bit of time on this element in my where that the time that I got this and what is that they all all the pieces of the inference procedure Europe there are 1 of the things that you know that the derivative of the 1 on the right things that you have to have been at the heart of the of the of the of the of the of the instances right that's now I probably won't hear areas this was 1 of the reasons as to why you think of any of but all you you all the all the all the data that you have something that is similar to that of the result of the universe not that so that's the only later in be you by the so I think it's something that's the only way to think about what the lack of this that is a candidate objects rather than the properties the model reveal the fact that you really want go out of the right to do this on the the other thing is also like what happens when you did the target of a specific and what happens so this sales and all that is the fundamental understanding of text and we don't have a minute it be did not follow also database-specific for the area here you have to have the the 1 the only 1 of the sources of the golden age of all right so you can that all the time and you need to create more often than today's world data that would you would roles noted this story the ones of course you you would have to locality the 1st and
then you were surrounded by a region of the world In the 1st stage of the game the you candidate in our and I think that they recommend actually comes from a lot of people this 1 mole of that so all architect and some so when you have different time apples master of groups and basically every don't know about each other and then how do you how do you make sure that resources are being shared shared really well and you them in it can actually will grow exponentially in because there is no restriction on it so you don't try and get as much fossils and memory as it can so the idea is that you don't run multiple passes what you do is you from a single process but then the instance itself is a multi-tenant inside but then we are able to control how much each you sort of alternate is is able to conceive from the candidate of his itself that's the grand design and that that actually is and what we are trying to exam this is the this is a picture of I want you to know that they're all the same things that you it so you have all this this is the 1 where the data that we need to talk to all of this and all that sort thing and it works the other way around as well to be honest like if you have multiple instances and when they want to do something there is available and memory or even CPU cost sitting there I believe can always use it but once you restrict them once it reaches that limit it cannot go any farther so that what is that and so it's not and they're perfect scenario but for most of these use cases we have no control that's what everyone in the world and of course there will be a trade off between 100 instances right vs. um you mean from a performance point of view or from a usability you words of resources such as is the only thing is over by there is definitely an overhead and running multiple instances there so that some of that's the thing here there that that's that's really straightforward there and you save a lot white raising the number of instances is not a straight away that I'm going to talk about and something that is not what is 1 of the biggest problem we have that we all have the same region was the result of that get get priority of people like me and you realize that you know that the true state of the state is an act of 18 thousand times and that the nodes in layer of the sun that's so this can also vary in the file you actually have a lot more about these being at the 1 way or the other of the of the of the goal of room for it to the that that actually is 1 of the discussions that would 1 of my colleagues started in the hackers system last year but pre 1 control and we got lazy and that again at hand and in so what counts as cited in some of the things there there so that but yeah I was a student that's that's not an easy thing to it's it's scientists that but it's easy to implement find state side difficult to define been by you 1 definition of what's important and what is not that that's benefit of Parkinson's story originally questions just doing so there is a case all the more what you know the law that he really was only a day period the start problems that you try to maximize this part of the you have the same rights as 1 of the things I want to to move things so that all of the of quality of the problem to have a thousand the reaction actually you that this is a new primarily back because of all these years ago all of this at the end getting populated area back all the exact same thing of all of this is the donor and some of their that actually you know the modifications required signaling everybody go all the way back here like all but when you you you you you you you you you know no and you can have multiple databases but and yes yes that's the way to think that you are not really the characters doesn't change them in exactly the same as the others what's what intend is how they consume the resources that and much of the world in the name of the I know it would be the other way around like you create endeavors and then you can assign users to read but those users can only see that others but if you want yes you can't restricted to a single user of small that can also be done with the with the policies that here you can do that as well In the last slide I have
this obviously what you are asking me how we limit the made sites and perceived as you mentioned we don't want to limit the instance size the instance has to go always it completely Shandong once it cannot fight anymore it's low so what we want to limit is the tables for actually the native me to put into that table space and and once it reaches the limit and uh you get a warning of something now again this can be done with seed groups and you can also implemented internally in kill as well but the difficulty comes in is OK so you have a table space of 20 or whatever it is you feel dig what happens next how you stress how you go do use below will create another table space of do what we do now is created a bigger table space and thing into it locking you a that ways of different classes you definitely don't want to do that and so that was the walls item will be at the and and then of course you know and who monitors that the other thing and do you create another demon to just make up refinements is like thus monotone bits tablespaces are about to run out on of space of to be filled in Finland and so all the all those questions come in and just like anything else and the implementation is fairly straightforward it is just the finding these policies and rules that what what happens when it goes out what happens when you we have dinner and hot how we monitor this of the difficult part 1 but if you want to restrict how much as stories you have of this is constant yeah it can do that already that put have the group around the table space there that's all I have actually moments any 1 of them we found that many of was the migration movements which we have about asking for your measurements and what have you have a lot of this sort of video was is moving the tables associated with tend to another machine they have a lot easier with the grows the need for us that to have this data database base right moving intelligent machine where there's more room in favor of troops but how would you just moved the table or how this migrated out of their and this instances is 1 of Canada's problem it's not I mean maybe not today and by some of the logic of applications with real world yeah so that that's that's where it becomes really really useful yet but that would be the logical and so on and the reason I policies with the logic of repetition of saw what happened so that initial snapshots that have more get all the news 1 is this of actually more I think you really think it would be that rate of the related to something like that and that they will have on your data on the Web in the USA and Europe where they freeze everything or you don't do something along the lines of what I should be able to only through the ages something that a lot of libraries that have you know that would work
on the analysis of this size this year
I have to refer to the same and so on and so on so you have different ideas about what OK to restart that was the government that you're starting with the same point and ending in in so you yourself are you doing here here what you have to near the unit is the reason for all the yeah that that that that is a valid point of but that's because of the the the the goal of of all doing what we think of it like it is 1 of the solutions is no later than when it comes up alliances you actually exactly do the opposite and try and I want that because you don't know the users excited unwanted that they want rolling upgrades in know and they don't want to just um whatsoever work on everything is that is the same the don't 1 multiple versions and things like that you don't really start a particle of time and it's just 1 day then they got plans they're sitting there and I'll go this is just falling and I don't want to shut down and stopped and things like that you want but there is still a lot of time the of the of the of the people each of the delta as a mechanism for this lays the wire breaks the whole wall continuing regional elections and everything in the going rate variability you century of it will matter and then all you work of the get the binary so I think that would be really cool which is what I mean the formula rolling with you know from falling on your PC whatever connecting the utilities in the form of agreed few new version of the search and so on there happened how we do it today is it that we already have enough plants and everything is and know and we get and the way we do rolling upgrades since it need to have this and intimately double proprietary things which holds onto those connections the selected set in advance of so that of of obligations don't get bumped off so we then switched to the standby over and then you the out regularly updated all the the the the the right now they were on the ball that yet to have you lose all that can walls of the data we love and by the way people is also but issues on how talk on Tuesday and that's what they're trying to what's well over there there there is a lot of the values of the troops in the pool that's all the more and on and so on and so on and on and on about it so we get I don't know what you think yeah this is the this is the way in is that all you have to think of it to what the 2 so we can talk about the lowest point of view that that is the number of years and this is the gradually interest in the middle of some continue had to back up because the next 1 has to start thanks so want to thank you
Betriebsmittelverwaltung
Kernel <Informatik>
Momentenproblem
Extrempunkt
Desintegration <Mathematik>
t-Test
Datenmanagement
Schreiben <Datenverarbeitung>
Extrempunkt
Information
Gesetz <Physik>
Raum-Zeit
Computeranimation
Netzwerktopologie
Softwarewartung
Computersicherheit
Translation <Mathematik>
Skript <Programm>
Betriebsmittelverwaltung
Auswahlaxiom
Gerade
Nominalskaliertes Merkmal
Kraftfahrzeugmechatroniker
Namensraum
Hardware
Sichtenkonzept
Kategorie <Mathematik>
Pay-TV
Computersicherheit
Kraft
Gebäude <Mathematik>
Stichprobe
Instantiierung
Bitrate
Biprodukt
Gruppenoperation
Web log
Dienst <Informatik>
Menge
Rechter Winkel
Festspeicher
Programmierumgebung
Tabelle <Informatik>
Fitnessfunktion
Instantiierung
Lesen <Datenverarbeitung>
Subtraktion
Multiplikation
Kontrollstruktur
Klasse <Mathematik>
Automatische Handlungsplanung
Online-Katalog
Mathematische Logik
Systemplattform
Datenhaltung
Virtuelle Maschine
Variable
Bildschirmmaske
Knotenmenge
Informationsmodellierung
Reelle Zahl
Spieltheorie
Datennetz
Migration <Informatik>
Programmbibliothek
Indexberechnung
Maßerweiterung
Konfigurationsraum
Normalvektor
Analysis
Tabelle <Informatik>
Raum-Zeit
Softwarewerkzeug
Systemplattform
Schlussregel
Elektronische Publikation
Binder <Informatik>
Einfache Genauigkeit
Generizität
Existenzsatz
Gamecontroller
Overhead <Kommunikationstechnik>
Wort <Informatik>
Partikelsystem
Dämon <Informatik>
Streuungsdiagramm
Resultante
TVD-Verfahren
Bit
Konfiguration <Informatik>
Reihenfolgeproblem
Punkt
Prozess <Physik>
Gemeinsamer Speicher
Minimierung
Versionsverwaltung
Gruppenkeim
Familie <Mathematik>
Regulärer Graph
Kartesische Koordinaten
Element <Mathematik>
Videokonferenz
Eins
Übergang
Kernel <Informatik>
Online-Katalog
Umwandlungsenthalpie
Einheit <Mathematik>
Datenmanagement
Font
Gruppentheorie
Uniforme Struktur
Hacker
Figurierte Zahl
Einflussgröße
Umwandlungsenthalpie
Nichtlinearer Operator
ATM
Interpretierer
Teilnehmerrechensystem
Zentrische Streckung
Installation <Informatik>
Benutzerfreundlichkeit
Datenhaltung
Reihe
Speicher <Informatik>
Globale Optimierung
Systemaufruf
Übergang
Programmierumgebung
Quellcode
Dateiformat
Frequenz
Dialekt
Konfiguration <Informatik>
Arithmetisches Mittel
Rechenschieber
Gruppenkeim
Information
Overhead <Kommunikationstechnik>
Zentraleinheit
Aggregatzustand
Sichtbarkeitsverfahren
Web Site
Ablöseblase
Implementierung
Zahlenbereich
Derivation <Algebra>
Sprachsynthese
Zentraleinheit
ROM <Informatik>
Ausdruck <Logik>
Physikalisches System
Benutzerbeteiligung
Multiplikation
Task
Verkehrsinformation
Mini-Disc
Stichprobenumfang
Inverser Limes
Strom <Mathematik>
Gleichmäßige Konvergenz
Speicher <Informatik>
Softwareentwickler
Grundraum
Hardware
Streuungsdiagramm
NP-hartes Problem
Einfach zusammenhängender Raum
Fundamentalsatz der Algebra
Graph
Namensraum
Validität
Einfache Genauigkeit
Mailing-Liste
Physikalisches System
Quick-Sort
Inverser Limes
Keller <Informatik>
Integral
Objekt <Kategorie>
Schlussfolgern
Minimalgrad
Flächeninhalt
Mereologie
Visualisierung
Mini-Disc
Innerer Punkt

Metadaten

Formale Metadaten

Titel Multi-Tenancy in PostgreSQL
Alternativer Titel Multi-tenancy in PostgreSQL
Serientitel PGCon 2015
Anzahl der Teile 29
Autor Shaji, Arul
Mitwirkende Crunchy Data Solutions (Support)
Lizenz CC-Namensnennung - Weitergabe unter gleichen Bedingungen 3.0 Unported:
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 und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben.
DOI 10.5446/19136
Herausgeber PGCon - PostgreSQL Conference for Users and Developers, Andrea Ross
Erscheinungsjahr 2015
Sprache Englisch
Produktionsort Ottawa, Canada

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract This talk is about the need of multi-tenancy in PostgreSQL, and the way to achieve multi-tenancy in PostgreSQL. What is a multi-tenant cluster? Why multi-tenant cluster is needed? PostgreSQL Provides multi-tenancy with the following - Shared Database, Shared Namespace - Separate Databases - Shared Database, Separate Namespace However multi-tenancy means more than this. - Issues with a multi-tenant cluster - What can be done and what can we do to make it easier. This talk will propose a multi tenanted architecture for PostgreSQL, to make it the database of choice in a cloud environment. Multi tenanted architecture is one of the key requirements for any software to be efficiently deployed in the cloud. As more and more databases are made available 'as-a-Service' in cloud offerings, it is necessary to take stock of the features in PostgreSQL to analyse how cloud friendly they are, especially for a multi-tenanted infrastructure. This talk will mainly focus on what functionalities are needed in PostgreSQL to make it truly cloud friendly. PostgreSQL needs to have the functionalities that will make it the database of choice for service providers in the cloud. This can be achieved within the current architecture of PostgreSQL by developing new features that will satisfy these requirements.

Ähnliche Filme

Loading...
Feedback