Tarantool an in memory Nosql database and execution grid

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Erkannte Entitäten
here's the house the the transfer of knowledge of the workings the of we would have on the use of the world and you must have
a lot of time and you can realize the holidays and the poem the beloved by him I decided it's but the be with them I have been in the minutes but as I had 15 minutes depend so and some of the slides as this that the shoulder thing and can you see from the back of the screen so um frontal is a project that exists for 5 years and it started just key-value database and right now it's something completely different than the beginning if it's so it's just the way interpreter so this is the 1st 3rd example of so it's just they're just the script and you can see it it and it's a little script and I'll just from so OK I managed to replace the edited while I was looking at the this is and you keyboard so expect some glitches so the basic idea is that we are a 100 % will compatible it's what it was it was a database in the beginning and end and is this old how would you actually configure a database our provide EPA eyes and we embedded tool in size and then erupting is around Soloway and we are just liminal and the interpreter and this brings us to the 2nd case I wanted to show is that but we're are not it done just that you know interpret with year law UX-compatible sickening is 90 stuff you can so this but was us on the field like a plane lower load jade or Toronto is on the same level so you might expect some glitches when you say the the fol lowing the system is low 5 2 but the fear of Toronto is it's so it's low 1 but otherwise we work very well with all the packages that you have will and then so this is an example I'm just here and then there is a bunch of packages they ship onward and it includes Jason which is lowest adjacent the HMM-based but we have a patch that so uh this an example where I was and trying to use the packages the CIA uh ensued some of them are just available and the 3rd example I wanted to show its a is a case for interactive mode so I am interactive mode is Justice something you've confined in a sort of an any model a language and in our case it's it's like that for you when you start rental without this script Anthony interact amounts of them be afraid just started and play with it did 1 difference between Toronto and other interpreters is it In our case we automatically of print uh it would be a result of the expression that you type and so it's a very easy thing since but the system for example you can instead G and it prints out a bunch of stuff so it's serializes the G 2 0 Young moment prints out to anyone has questions about energies the global environment in water OK so some other packages we have on word 1 of them is box the and and the boxes our database so this is how we're upping eroded the built the ship the databases so as just a package is always available as a global 1 and so in the in the important things for R packages the road be reach the is that all we often provide our own versions of some of the standard packages warm sinister and like socket and the reason is that we are not just the low interpreter but it said so you're aware of not GS Sorel lead and things like that so we have on an event machine on board and most of our of our practice serum deal so that they're even machine aware for example we have filed input output the practice called the and the this input power output is completely non-blocking so you can the it you can open a file and the created and even node the law itself a single-threaded this is is it is all happening in a thread pool which is available to you at the end of and machine which is
running behind the scenes said to just give you a better idea what the repayment mission is is like we unlike Asynchronous callback based environments we are and we integrate man machine through fiber environment so through lightweight threads which always on the same physical network that are operating system thread so we call fibers and their this is my for the of 2nd example this demonstrates the the use of fibers so so if you if you are an energy as you or a lead user are this should be familiar to you what would like what what power you get with this approach but Our approach to FHE to and I known event-based programming is that we don't use callbacks the code is completely lay near the fiber gives you this transparency that you recency run like a thread but what would you rather than institution convex like a simple program in any any kind of problem but whenever anything any an event happens through wait for an event you your fibers and implicitly taken off from the chandelier In another fiber takes so execution so here are you in this example I just create very simple divide fiber and the fibers exchange messages by channels channels is so it is very much like goal channels this is this is a star implementation for a long and this and this this fiber just in an infinite loop just getting the message and a channel of the messages exited so it's it's existing in other words as being from the message and this last line is over is a way to start the interactive mode from discrete so I'm understanding this last line to script those speed up things and and I know I don't have to type the script in and I can I can just a valid escaped and then continue name in interactive mode so for example now we started the head but evaluated and I will now try to create a fiber so this is the fiber package an interactive mode I have to global variables in mice keep they aired the f is the function and channel as the channel general variable and now I create the fiber in the channel the and remember this is all happening the SSH to my machines and so I created this fiber and it brings here the fibrous suspended that this identifier I will cancel the fiber Nike here fiber killed I don't want that this fiber around so I killed it and and then I will create a new 1 and you see the idea is to do this is a new idea so this this this fiber if you remember the code is waiting for a message for us and that generalize the variable that we are going to use the send a message and you see the message that sound we get through back from the channel put function and this is a print from the fiber so I will send another message and there you can see that there no the 5 that the message announcing synonyms message the fiber so that basically you can see that there is a background fiber and there's an interactive fibers and they change messages issues with each other so we provide this programming model to for the user and this 5 fiber is dead now so the next example I'm getting this is already the example of and the database and here are the data with just a package put the you can't use that package and what I'm in the database is something persistent so to begin using it are you need to bootstrap but and the whole thing with straps automatically so I will this CFG configure function is something that we'll either bootstrap the database or loaded from existing state so when I when I start this kid for the 1st time let me check that they don't have any any stuff and 5 it the yeah it but when they started for the 1st time Mr. you can see that here it saying like I'm creating a snapshot here and this is a database snapshot already a snapshot is something that contains our initial are tables like several existing tables and progress or my sequel and so it creates an empty snapshot to adjust the system Table science ready to go when they started for another time it is would you from an existing snapshot so now
that I have the database configure it I will try to create some stable physical spaces and there is a column spaces we have a fun in data model which is sort of bridges the gap between document databases and the relational databases so we can store you can store Jason them or you can use them as a relational tables so I will not create 1 table as you can see on diffusion lower so it's a based on their
um which you see here is that I be IID transactional so this change good peek into the write-ahead log and then on the table got
created it's has some properties and the storage and particular now I will define and indexes and yeah so
that this stable can receive data now really I will try to load data and I have little time and I be um so were fluent thinking I will try to load debate cannot I have prepared file which will open and so and so I will iterate over the lines of files bonded this it is a comma-separated well is in the file and so they can to my table would save the force from Hey it here soaps good obtained so are just the desire lines in the file which I think I answered so of the ideal flow of this whole thing is that your databases programmable in law and at the same time it's a just a server so you can connect them so I I I have already configured an hourly reconfigure it so that it sir it's listen in on on a socket a case that's already alluded worried and probably now and now it's up to some of the problem the sum of the Toronto running yeah better now um the so it's listen suck at let me just show you the idea of hollow quick performance database is I I'm having something prepared here for a benchmark it's a I am I define a working directory lists imported and then our our listen socket it's a prison space there is often the case and the the test control soils prepare some grounds for users and I granted the fold using the gas the access to the to the entire universe here helps uh
so books once is our way of the term would stop in some some actions which are only happen once in the clusters so we have lead to support replication and Our in replicated environment you want some changes like system spaces with creation of space so grants to happen only once you don't want a repeat this actions in every replicon because of all the actions are already done and my for example so this is where I propose OK i 93 do it because I so this is the benchmark until we wrote and the it's now in certain date it's 100 couples 100 by couples into space I created in the streets so you can see that the average performance of like 450 thousand requests per 2nd for network network and that of performance said to be actually a bit more our fair to this so this this that high performance is achieved by his intemperance and then every thread descent send in a question budgets about is a thousand requests if I reduced the size in bytes the performance actually goes down because it's I there is more you know ping-pong setting for a request a lot of prior to that of a yes you can you can model we like you you you used the and times and you know what is all of that growth so far as to what the buyer would be of the of the of the existed in I that kind of thing we can we can do this also don't yet so it was just a complete and know what time it was the most time I used in the middle of the here and then you have to have a form of you we would do you might use
Kontextfreie Grammatik
Konvexer Körper
Gesetz <Physik>
Service provider
Metropolitan area network
Arithmetischer Ausdruck
Skript <Programm>
Urbild <Mathematik>
Funktion <Mathematik>
Metropolitan area network
Lineares Funktional
Kontextbezogenes System
Funktion <Mathematik>
Faser <Mathematik>
Projektive Ebene
Tabelle <Informatik>
Regulärer Ausdruck
Interaktives Fernsehen
Ereignisgesteuerte Programmierung
ROM <Informatik>
Demoszene <Programmierung>
Virtuelle Maschine
Nominalskaliertes Merkmal
Arithmetische Folge
Leistung <Physik>
Physikalisches System
Elektronische Publikation
Patch <Software>
Serielle Schnittstelle
Wort <Informatik>
Prädikatenlogik erster Stufe
Relationale Datenbank
Bridge <Kommunikationstechnik>
ROM <Informatik>
Data Mining
Tabelle <Informatik>
Metropolitan area network
Gruppe <Mathematik>
Kategorie <Mathematik>
Automatische Indexierung
Güte der Anpassung
ROM <Informatik>
Wiederherstellung <Informatik>
Tabelle <Informatik>
Offene Menge
Stabilitätstheorie <Logik>
Gewichtete Summe
ROM <Informatik>
Gesetz <Physik>
Metropolitan area network
Cluster <Rechnernetz>
Pfad <Graphentheorie>
Physikalisches System
Elektronische Publikation
Data Mining
Tabelle <Informatik>


Formale Metadaten

Titel Tarantool an in memory Nosql database and execution grid
Serientitel FOSDEM 2016
Teil 102
Anzahl der Teile 110
Autor Osipov, Konstantin
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 10.5446/31016
Herausgeber FOSDEM VZW
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik

Ähnliche Filme