Merken

ORCA Query Optimization as a Service

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
and thank you for using the interest of the people a to 2 on 1 of the more this is there's no way we can say that we can do with the people have to walk and 1 of the most aspire to part would be optimal and so on and talking about latest addition to the open-source from computer of the Protestant work this essentially universal optimize the teacher in a useful way outside of the project that we are using 1 of the the limit of the a lot of you have you been telling of so again since the room and prices the easiest thing like that and then that I was explaining analyze it and again get a better school so I went I require optimization is because of the way they're all day but we might get back to this so working on it and the thought of this type of time in the college re-optimization for as service I don't think about exactly that but the it's a very unique approach where urbanization that we think about a In these itself and 20 minutes on this on a little bit about what is gravitational waves you care about them interviews or they're belongs to the internal we're going to go through the process of pairing interaction it had transformation in the work together of them my talk about where things so why do we want 21 was in the of so organization through it takes a expression of stories that can be a single operating on a MapReduce context the and it turns it into an execution would have should the data so you can the but the problem is the way how fast the growth is not the way bigger than our ability to process so we need to use what we have today the process of the particle were and historically every relationship with its own way you think about the idea that they have weights 16 of which is a lot of overhead maintain and actually 1 of the things in the data is just the last day of exams the very complex and the codomain in so this more water comes in there's and buzz words but I want to do as a worker is a modular approach to query optimization of when we're thinking about its architecture was making friends and if you think about the psychological plexity of program the existing posters where this work order in 6 to 10 times of lower on that scale making it much easier to maintain it at it's also extends to the way that we give you add something to you can add unity transformation on operator for another database variables used in support of all this is the most interesting that is outside the core of the database and you can be used for multiple data so it's used today in more than 1 and we think that used in a lot of it what's interesting about this makes it perfect on the test for reactions so you think you can add something to this field or is a great place a chance to to try your ideas but it's more than just on the test but it's actually used in the production rate databases and in the customer they're guaranteed but also the rest of the police
chief of across the entire DCS worried transfer is already is 5 times faster than you were not using it actually falls back to the late the light of these down near the regression of talk about later
and has a Friday of the source 6 years are being given to the world that is we think the way that we can make the letters but they were released so what makes it so he wouldn't think of work actually doing it but I think this is what it looks like the 2nd 2nd so some key features we do really sort of petition elimination they interact grabbing data from a real meeting of the nests of queries we tend to have the present in the reversal theories were necessary but we with killing water and power renewing the sorting but it's really aware of the of analogously to understand life so for any further this difference between 2 logical operator and physical operators this is really important understanding how that work so a lot of operator is something like a joint so this is just the of logical way of thinking what needs to happen to the data for 2 the the physical operator in the algorithm it's actually going to make that operation happened in the so so a lot of operator that actually is the algorithm that went so a lot this is what will do look through a couple steps where the preprocessing of the logical representation word the end of this 1 logical for a way of turning it into a physical implementation so you couple transformation pick 1 Due to the physical properties and then the best the of of that what is something that we do preprocessing processing doing what want logical as a example that is kind of them so we have something where positive with that part of the tree so they were not thinking all the data in the frame of the tree or select will mean that the part of something really simple but every single 1 of these 2 but we do all that really is the preprocessing we do not hold of logical transformations and then at every single 1 the logical transformation plants we actually implement its physical uh place and then we can use costs all of the so if we were going to say to yes word 1 we think you 1 union ways of actually achieving that is independent the cluster and a the metadata surroundings I will choose the right way that comes so the couple biological fascinating talk about how to draw an ordering of the the standard 10 in that the analogy of a whole host of this is just the a 77 more than of playing the lottery with so I'm like this but there's a lot of you can and the reason that a lot of the margin you know you have a deal of like all I think could actually compute this this 1 3 different 1 and when we're developing these that's what we did you just call there's an idea that had and then for the way that it's architected distress so uses something called the itself the community that interest so it the following thing was the quality of its inside the whole idea that the in the and in the in what we mean when of at of the areas where the new face of the itself out we get the idea of representation of of we allocates memory you're doing we look at the operators necessary you transformations we have a vibrant effort cardinality estimation cost model so so and once we have a good plan we just send it back to the database in it so it communicates over the so something look at all
the uh as in a word of we care on so we're the inca so what I wanna do now is our and transformation together but it is I have an idea so an idea and that's displays an aggregated into a pair of a local or global this is the the specific operations In other words these site work but he heard a lot of the databases so the question as I don't know better to calculate a dealer if we have a scalar variable can Harold we should help on 1 of the segments and then sense that people to the master or if we should send all the data to the master students have been giving up north of the no idea so I wanna know transformation that's worker will calculated to figure out if we actually what would a and the the plan that went actually gather all those tuples in the history of back the so an example is we're going to select the sum of column c from data to improve by calling the United States it's supposed to be that the with a cool I know where did you filed is where the source files are and the new added destination we part so it follows that pretty basic pattern so we have in these transformations so commercial estimation is the goal of this form so little time and in the same way that this personality the to use Hungarian Notation 108 haters out of its hole to POS is something that I want you to think rationally on top of the operating system uses for a bunch of different things may also sorted so if I call you is the left it has to be going deliverable by meaning that they must that we will use all my Gilead plus transform so that's a good thing so this expression of many of the things you a logical and operator in that the context of the global and operator for the class and the next thing is the relations child and give it very so that matches to the transformation of the at the back of the unit area have been the subject of the of the original child from which is the the the skin and there are scary help so this is the work of the the things in the actual trajectory is triggers that say OK to what was literally this introductory but basically inquiry during this those thing so worker has the function will check if we we to the stretch so are all the conditions mentioned previously if we call this version functional they will actually go ahead recently where transformation easy way of so we pass support in this function is that from the beginning of the contents so that the over all of transformation right of media said the result of the transformation fact so work that the larger population were just doing what he's dead transformation of in the tree and that expression that we wrote to reduce the available child the and that will act and then we have registered with the battery the so there's a of this has got variable and we use a memory and then we register our the forms political ad hoc that work will go through and I see that actually the fire and so what can you get a lot of work so building core
everywhere the turns of lot life it's about finding balance so right now the it doesn't really do the job and it's these long running he he worries the table vary over right now we're trying to figure out how to proceed with the
right word word on the job of extracted feature complete there's a lot of things that the Fourier external external parameters were added in the next section of things that you ladies he was appointed so what's so are all middle of this project is to 1 day we place have because this but it's not working the fact that the theory a long way away but we think we can get there because the the number of people actually but I think you can think of this as class so you I have a much longer of line through the if you want to start working on I work on there's no actual distinct distinguished in the physical logical operators the the the database knows about but it's not all of this in the region inside or go through the person and that in 1 of involves I have to live in the real world rather than you know a lot of the databases used reason that you get the time that tells a talk about of the dead the public difficult tracker that we used to track of all of our work so you can get their on a single working with data want more details and into the economy the topic of presentation was in the the whole thing based on their features this time through the they the top top and when and also of the fact that so there's a very good database PA so what did I do it when we get to work now I think this is major there's all the data the the the the the experiences of so we down the metadata from the that all goes to the use of a of the of the of the whole thing and because of rules and we have and you wrote about has kept that has cast in the form of that the metadata of the use the running costs so you can get a better sense of the in of the very top of the actual cluster information like through the clustering and more and more the generally think of this size and form of these of effect so we try again the stats is on the bottom and through those that we have that that's would be across so for example a lot of times you were not getting the typical point in the sense that the means to the of data statistics and often those that were going the because this is the this is in keeping the problem where you need 1 the guarantee of loans that all the other end we have reviewed stats and can't find and I told you the kid you theories of all of the time on here so do you know that you're actually that's not something that we run the lead something about like that did media really run a so what people will know what other cultures so what is that the peoples and probably the that the people of the same thing you can think of functions would be the task
Retrievalsprache
Bit
Gewicht <Mathematik>
Prozess <Physik>
Selbst organisierendes System
Wasserdampftafel
Minimierung
Interaktives Fernsehen
Dienst <Informatik>
Transformation <Mathematik>
Computer
Computeranimation
Eins
Arithmetischer Ausdruck
Variable
Multiplikation
Web Services
Datentyp
Inverser Limes
Optimierung
Softwaretest
Addition
Nichtlinearer Operator
Zentrische Streckung
Produktion <Informatik>
Open Source
Datenhaltung
Wertevorrat
Kontextbezogenes System
Datenfeld
Mereologie
Projektive Ebene
Speicherabzug
Wort <Informatik>
Computerarchitektur
Partikelsystem
Ordnung <Mathematik>
Gravitationswelle
Randverteilung
Offene Menge
Retrievalsprache
Subtraktion
Prozess <Physik>
Rahmenproblem
Wasserdampftafel
Selbstrepräsentation
Physikalismus
Automatische Handlungsplanung
Implementierung
Wärmeübergang
Dienst <Informatik>
Eliminationsverfahren
Transformation <Mathematik>
Kombinatorische Gruppentheorie
Systemzusammenbruch
Mathematische Logik
Physikalische Theorie
Netzwerktopologie
Metadaten
Informationsmodellierung
Algorithmus
Reverse Engineering
Lineare Regression
Addition
Ganze Funktion
Analogieschluss
Leistung <Physik>
Schätzwert
Nichtlinearer Operator
Videospiel
Präprozessor
Kategorie <Mathematik>
Datenhaltung
Abfrage
Quellcode
Quick-Sort
Verbandstheorie
Flächeninhalt
Festspeicher
Mereologie
Wort <Informatik>
Eindeutigkeit
Resultante
Retrievalsprache
Web Site
Subtraktion
Gewichtete Summe
Datensichtgerät
Klasse <Mathematik>
n-Tupel
Automatische Handlungsplanung
t-Test
Versionsverwaltung
Summengleichung
Dienst <Informatik>
Transformation <Mathematik>
Trajektorie <Mathematik>
Netzwerktopologie
Zahlensystem
Variable
Bildschirmmaske
Arithmetischer Ausdruck
Einheit <Mathematik>
Prozess <Informatik>
Netzbetriebssystem
Mustersprache
Addition
Hilfesystem
Schätzwert
Umwandlungsenthalpie
Videospiel
Nichtlinearer Operator
Lineares Funktional
Datenhaltung
Relativitätstheorie
Quellcode
Kontextbezogenes System
Summengleichung
Diskrete-Elemente-Methode
Flächeninhalt
Rechter Winkel
Konditionszahl
Festspeicher
Hypermedia
Speicherabzug
Wort <Informatik>
Tabelle <Informatik>
Retrievalsprache
Punkt
Klasse <Mathematik>
Zahlenbereich
Dienst <Informatik>
Kombinatorische Gruppentheorie
Physikalische Theorie
Task
Metadaten
Bildschirmmaske
Prozess <Informatik>
Cluster <Rechnernetz>
Gerade
Soundverarbeitung
Lineares Funktional
Nichtlinearer Operator
Parametersystem
Statistik
Vervollständigung <Mathematik>
Datenhaltung
Güte der Anpassung
Einfache Genauigkeit
Schlussregel
Statistische Analyse
Fourier-Entwicklung
Arithmetisches Mittel
Hypermedia
Wort <Informatik>
Garbentheorie
Projektive Ebene
Information
Speicherabzug
Computeranimation

Metadaten

Formale Metadaten

Titel ORCA Query Optimization as a Service
Serientitel FOSDEM 2016
Teil 31
Anzahl der Teile 110
Autor Huddy, Addison
Lizenz CC-Namensnennung 2.0 Belgien:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen 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 10.5446/30943
Herausgeber FOSDEM VZW
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik

Ähnliche Filme

Loading...