Merken

Systems Integration: The OpenStack success story

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
OK with welcome come to the 2nd talk in this session given by further pack or on all the states thank you both of those words yeah they bombing FIL
around today so I pull my talk is about system integration it's not actually at 100 % related topics that are
mostly related to integrated system at each other and all use of a 2nd
example because most of the methods of presenting you are being used by Open 2nd itself integrated services that we so that's me that's my tutor handler for much everything you want to know about me out there in the Internet or of the I want you to know I worked for Red Hat and part of you community are used community of a bunch of really great people working together to make a precise amazing on on road base distribution what it thinks about me and they're going to go through those the ones I will mention his common argument Summer of Code and OPW mentor and I want to mention this because I really believe it's very important and if you have spare time in your data you want to mentor people please sign up winning so a tree before we go through the methods are that you would use to integrate systems or let's 1st define a little bit what system integration it was system integration these basically what you do when you have a set of subsystems and you want to make them work together toward a common goal or scope right so all the methods and technologies and strategies that you would use to make those system work together and to get to with the goal our is what we call system integration is putting a very simple way of there a whole bunch of different definitions of system integration you can integrate system systems are known as source of you could use hardware systems use many of the existing system so it's this is a very generic term that you would use to say they have a set of subsystems working together are for a single calls so there are many different ways of generic strategies to integrate systems but these are the 3 that I will present very briefly and we will dive a little bit more and the last 1 by so vertical integration for doing vertical integration means or basically looks like not too small graph appears it looks like let's start with you have a set of systems that you will talk to system the system the system that about which of these is the better and it is a strong base that each subsystem features and what you need from them so you have a web service that integrates to your database and then you have 2 systems are working together or you will have your authentication series and then you have your order services below and you would make this service with the real features talk to your authentication services and you are integrated those services radical this decide integration what schools so integration because it's supposed to be like sorbitol like spaghetti integration because all services know what all the services do and the old together and they do that in that case by case basis so you
have service a than it's something from service b but before doing that it's up to service e because it is something from services before going into service being so it's quite a mess
on use cases for these the kind of them all but it's very risky and very very error prone and this 1st year of very high rates of not having a contract when you're talking to Windows services of talking together by not having a contract basically means that you don't know what you're going to get back at it you don't know when something goes wrong when you get something from service c that that the service the but service it's expecting something different and of the service he was updated on and the other method that we're going to die for Livermore today's horizontal integration by and a it's based on the service bus and Service
Bus at call the communication I'm I don't like to use the term message bars here because it doesn't matter to me it's not about messaging itself but about making those services communicate together through the same
box so you have serious cases the insurgency and the all communicates through these communication was sending I the messages or just the data that acid that would make out the whole thing to work through this communication because so that in a little bit more of horizontal integration
from an application's point of view so how would you make all these so imagine you have
a set of obligations that you want to make them work together you need to have these communication but so you have to come up with an India with the technology that you would use to make them talk together wanted to prison like 4 different methods for for different methods to make those divisions that together and desire not new methods that are actually been around for a long time many people use them and they actually don't know that they are basically the witnesses summoned the doctor don't know what the whole thing they're doing is the this and which 1 of these cases are good for very sorry but each 1 of these methods are reduced for a very specific cases of some of them are more generic and others are more specific than the 1st 1 is the story in the first one is files of closing strolling sprawling that all this way to integrate the different services the picture for long time people used to open a file or get a file descriptor put something in there and having another application within the same piece of hardware reading out of the a that acid that you will use to something so some people we use follows 2 of the we use files as a messaging like you we use are evident here messaging broken but it's good for very specific cases and try not to use it is good for cases like embedded systems in and better assist you want have rubbing in the running for sure so if you have very limited hardware and their limited sensory memory you probably want to use something that's really chief files all the fully accessing the full system has a coast and it has a higher risk in terms of security and reliability but it works for you worked very well for embedded systems on we used to use these powerful of we used to use files in a perspective have some kind of being server distributed along for some time but many things went wrong with that so don't do it we are now working on the way to have a distributor lots but that's 1 of the cases where for example that we use files and when we moved away from them because like it's a very very good for hardware that has very limited resources but if you can afford something more it's Mrs. people who want to databases does this is higher probably 1 of the most common 1 and anyway all these statements are based on my own experience i don't have really actual data that prove that this and this is the most common over the files are the most all this so this is all like this might be as my own research has some databases are probably the most common way to integrate services they are assigned to nos data wise what that means is that when you have a message into 1 another service to get that message you would just stored in the database you're done with it so the produces so the message in the database the producer all with the message and then a consumer eventually will get that they out of the database and would do something with it but there is a really great for foreign states and I'm saying this is probably the most common 1 because I like most of the web services of their like I couldn't think about what service that does not rely on a database and you want to scale your web service you definitely have most probably having single out for the whole thing and you have civil service is talking to the database and getting data out of the right and there great bravery great for states and the way we use the single beside universe that most of the series of pulled the biggest services are halved I have have been split in in several smaller services so you know that for example nobody's the computer how many of you know them or have wasn't so know is this service that is responsible to spawning new instances of emissions that Pr always like is is used to for a the worst but and Nova has 3 sub services or what has it has many more than that but that the main the main services you need from Nova like 3 or 4 services and you have the EPA cities you have to compute the the the compute node and you have been steadily and you have conducted that gets messages and sourced everything that it so when they when a request for a new instance comes into the API series a new record will be creating the database and then a message will be sent to the stable that would then will talk to the object to the novel compute notes on the new little machine so would never compute doesn't get the data of these new users that was requested out of the database by exposing at machine and when the dimension is running it will update the state of the victim using that age of admission is running and will what they did in the diocese integration in this rule is small scale and that's the way you you could use of assisting so that's why I say that some they're probably the most common way to the resistance and and problem many people don't know that there are 2 integrated system by using databases stock so data
I hate you perhaps for so there's any of
you have any questions about feel free to interrupt me if you have questions but clear evidence stuff but possible so messaging would it
be like what it means by messaging here is not the
broker is known and be and is not on the specific technology that allows you to send that message from point a to point B what IBM about but I mean about messaging here is the message itself but the messages that you need to send data from point a to point B whatever method you used to send a message from falling into brain be up the benefits of using messaging is that is loosely
coupled and it that's way more complexity because maybe musical bolded means that you don't have a contract on the message so the series and can send a message to set to service b the series has on it but they did idea of what it's going to get and when you want to do with that message get more complexity because if you don't know for the message may look like you pull will have some passing arose all type errors or whatever depend on the language and what you want to do with that message but some benefits those that I've been loosely coupled you can say I'll send this message and mean ever gets his message can do whatever it wants with and 1 of the places where we use these kind of all messaging or loosely coupled contract is in similar but where the social media would plug into the notifications stream of OpenStack and will get all the notifications of what's happening in your infrastructure if you spontaneous to machine notification will sent so somebody gets its processes and thus something with it creates events create starts and allows you to build users based on what you've done some 1 thing about messaging is that may depend on Mr. Schroeder's and transformations so we use messaging and you want to send a message from from point B 2 . se but you have to go 1st of all point b you will need a point be some kind of logic for of technology that will allow you to run that misses the point C and you would do that based on the message information itself you so you have to know what to do with it you have to try to pass it and get information out of it to know where the message has to go and this is something that no muscular for example just like it doesn't get notifications gets are an message and will go to that later but it gets a new message passes in and it tries to get another computer know that will do the work for it and it will send that message computer and it does that by using some food and our ability and on all the important information but that's not to but this and the director of they're very easy they're very cheap but they add complexity to your and the last method I want to presented a is on the and RPC but in that sense for remote procedure calls it was for all interviews pretty much by danger price for when about system integrators wanted to integrate systems for the customers and they would go and use RPC calls to to do that and obstacles and provided the weight works is you will send up for modern messages you have a constant and the messages to point b . b would do that and it's called Remote Procedure Call because your basic calling a remote function just by sending a message you will say call this function passing arguments to function and give the result back is the most used above method and I do have image for this the message channel may vary you can use databases message broker so like I say I'm not talking about about here I'm not talking about the the method you would use to send the message from point a to point B in the past decades there we use message brokers to this article purposes and 1 of them the drawbacks but it's actually something required for a perceived these studies tightly coupled so you have a problem you have to invent something you have to agree on the contract when you send a message from point a to point B. because you want to call the function that you know it's point B you have to pass some arguments that function you want to get a real life so you have to know what are you going to get by going have to know where you have to send to the point be call that function so this will technical couple you will need like design your own particle to this 1 but you know that it's it's really common it's very useful for doing that kind of remote calling function thing but you have your benefits and drawbacks taken from this and that is the case this pretty much a high-level overview of of hope that works in terms of of disintegration is based on shared nothing architecture of the notion of an architecture is it's basically and reversible weights services working together but not sharing anything and the natural anything I mean they don't share of memory space on your box they don't share processes the pizza and and and other resources like they can live together on the same but but they would share the same resources they all have their own space within that that so every service knows very few things about all the services and by and with that we we managed to get and keep all the services vary as a letter from each other which is really good if you want to integrate systems together you wonder services to be independent wonder services to be isolated phrase each other and if something happens to 1 of your services you would see the W 1 you're all the services to still be utilized and be able to to work on the on top of all the services your system so we use that a basis for inter service communication like no was nobody can I will say storing new instance record with boating state and then of you will that state and we use Sarkozy for inter service communication when no weight guy gets a new instance request will send an artist in measures of message the skill and the scalar will get the message and they will send another artist the message to the computer knows that would then would be the retention but we use messages for cross service communication and Byron mentioned is the ways it works is that services when something happened without the 2nd services will generate notifications Notification they will send it to some specific topic in in the broker that all the services can just wanting to and get messages out of it and they can do something with those messages on so this all this and realize a lot of brokers and and strongly but right now 1 of the common tools to to integrate services and in in many employment like to say a few things about brokers and and the technology that you could use or how you could do integration based on on propose I think people are just using technologies like Mr. problems so this is the 1st thing I want to say is the scaling properties will hard you may even read or heard something like brokers scaling is already fixed and you can scare reverent you I'm sorry that's all like that doesn't work that way and there's a lot of the accommodation yes there's some explanations how you can do yes there some demos that people has done it yet when you get it to big scales because the 1 exoskeleton brokers is hot because synchronizing messages between different nodes of your broken that is heavily reddened and heavily written on but it's really hard and it doesn't work that way another thing is that brokers need a lot of memory but it depend so your use case if you don't have many messages traveling around your system pull 1 used a lot of memory but if you have the deployment your broker is definitely going to use a lot of memory and you really depends on how
fast you right to it and how fast you read from it if you write really fast and you really as fast as you read but also received various sizes you write your broker will call we use all this money did the memory of the training will be truly and stable but if you have more rights than rates your broker will use a of memory brokers need a lot of storage and if you want to have a durable cues and and you have your messages to stick around if something that happens you probably will use a give if you use a durable Q your broker will have to write everything on this because if the broken goes down have to start from somewhere right so will read all their messages out of that wherever database over or a storage system it is using and it will make those messages available again so again if you have a lot of rate of rights and they're not as many reads as you have rights you're of your broker and your broker will use a lot of stuff that was a look at the time and it says 9 minutes because like the brothers went down there was like more already but so like this is I've been writing about brokers for a bit all like to say something about those of if you're going to use brokers or any messaging technology prefer federation instead of centralization but what he made ideas if you have a centralized broker and you want to scale the broker and the broker goes out the on your system is off the and you will have a Chadian or what you was like you have you want to scale brokering you want have been replicated in all of those kind of things but if you prefer federation instead of centralization you will have a whole bunch of knows there like language brokers that workers and if they go down you will probably set up and you want and you want relied on a single broker based in the middle of your system of assisting older messages 1 way to do that it relying on major don't but no and you don't we know and publisher you most of you are familiar with QP itself current latest version of the p protocol being used by rubbing Q most of the brokers the same to be 0 . 10 an Open Cube open tennis not the and many brokers have implemented in different ways of whereas if you don't we know is actually asunder and it dictates how messages would go from point a to but 2 . so high you can share the same messages from between 2 beers and thank you want Windows peer based in the message bases where my daddy dictates a bit spies how message will travel from a point to another point but it also in the specification there's also explanation of how you would do that with an intermediate broken so it doesn't say that you have to have completely federated system you could also have a broken annual that's capable to speak and there not so into bumping it's all about messages and how those messages will go from point a to point B but and if you want to scale it up and have more writing the intelligence service being system you could use something like the dispatch that will allow you to create new rules the centers messages from from between your services as you would do with using routing caves in and get point so I know many people we know you don't have it changes you have accused you binding rules and you have very in a could only know you just have messages and you have links in every link spacecraft connection but 1 of their peers in your system so after having said all that about methods to integrate systems and technologies that you could use and protocols and all that stuff all like to give you some tips and tricks about our system integration this is mostly based on are experiencing in the 2nd in the on season for most transmission protocol matters so much broken the took another little was later like them at the UTP against TCP undergo but I probably higher level like whether you want to use a protocol that's TCP why do you want to use it to be where you want to use some of the IPC Pericles whatsoever on transmission verticals matter they depend on the broccoli choose you have some it's circles on on what goes on your messages and and transmission of of your messages so be aware of that a dependent on your use case make sure you choose the best particle 40 years each year this use versions for your Y particles of if you choose to use PC to integrate your system so you probably will have to agree on a protocol and you probably have to define that political by yourself something that has been arrivals that for a long time the version of this proposed so when you define your broccoli polar say my broker Lisa dictionary there will always stand between services in the dictionary has a key that is called function and that function is the that keep the value of the is actually functioning and then I'll have gotten KKT keywords and in the dictionary and that will be the volume to pass to the arguments and keyword arguments of my function but then you want to be that for a you say I want to also specified the return type 1 from that function and if you have you have your system deployed and when you want to make a change to your protocol you can do that but if you have version and you will probably have to cheer all services down and then up again when I was your data the program because some server gets some services service B gets a message on a PC messages with a is an RPC format into the recognizable pull fail is that if you have version and you can do rolling updates and assistant by restoring services 1 at a time and update those services so you have any doubt version is not just useful for a great also useful for all backward compatibility if you do a change in the change the result to be really bad you can go back to your previous version you still have your services that used to work at that session keep everything explicit on what I have a really nice quote that I got from Jeff haulage talk about the riot conference he basically said in a distributed system having implicit things is the best way to fuck yourself and that's really true like if you haven't implicit things happening in your system like you send a message you don't like an artist message and you don't agree on the contract of for the message you will probably face some some real issues that you don't do DSpace happen on so keep everything explicit even if it is more verbose even if you need more code if you need more nodes running that's fine just keep everything explicit
because when something bad happens you will know what it is you know how to the body and how to fix it most of the time can you don't have to make up and get can repeated like he's asking for an example of an implicit 0 I that 1 of the OpenCyc issues of and you will see light but for a long time a couple of ways to limited use like similar users messages are begin get messages out of their on notifications of that and there were some implicit fields being sent by some services and those fields wearing black were sent by all the domestic services so submitted you know about that and when there was a case where it failed when you've got says messages but the thing is that it was before the release so it could be fixed but anyways you don't you don't need to like something that you want to cater it to city is how you distribute but hello where's your notes are running and would write 1 nodes can run along sigh with other notes you don't want to have all nodes run on the sensory server so if you keep your accident you distribution very explicit and even in the way use different services it will be easier for you to estimating the scale and and how to use as a good example of this is no itself again how Nova has a new like a service and has another a service so you're getting a lot of are EPA request you will get a lot of messages and you're still if you still is on the lot of pressure you can add more still asleep and I will do that you can scale them horizontally varies so the way you distributed services in terms of of of code like useful you have an idea there a scalar service conductor service and compute servers it's another way to being explicitly how you're distributors essential like this and the contract and then using the word contract a lot today if design the contract you don't have to to like you know when web service these expecting you to stand and service these expected you to send something so service beacon can are wrong it's a set of observations before running anything in there will be a reply back with if some of those requirements are not met so far far that's pretty another way are we need to reduce system and you want to services to talk to them to talk to each other you have a contract between like pretty much like Europe account manager and yourself what you have a country with a you know where he is expected to do when you pay something that you wanted to get all the receipts and give them give the research and a you know when you get there is said to him he will do something with it and you will pay him for that service and willing be he will expect you to pay for his service right so you have a contract the same thing happens with services when you said that requests service the service this expecting something to you from you you know what it is that the service is expected so you will send that if you go mentors requirements you will replace that with an error and mean and if you send all the requirements right expect something back from it if you don't get when you're expecting you can just call back again silicate this was not what I was expecting so please give or whatever these about them because there was formerly known by most of you it's what was introduced by a folded programming language and and this is part of the coding style of the language itself get services isolated as much as possible on but it should not an architecture use is very useful when you want to keep URIs distributed systems by from fillers like and smell like implement from failures but if 1 of your services goes down and it's isolated from all your other services people we can just run another 1 somewhere else and just make to so could then isolated keep and you get your services various very stupid if you can bomb and I'm not talking about micro service architecture and have been like thousands of micro-services in just 1 of little function things but keep then isolated and very focused and context on what they have to avoid dependency cycles between services so I would recommend using this integration method it's really messy and when something goes wrong you it's very difficult to the bottom so avoid having the Henderson cycles between your services if you can have a serious because you can send messages the rate make sure you don't depend on the both services on eventually took to get something done models not testing if you have a distributed system you probably want to test if you want to test you will say hey the reasons which it says by mocking what I'm expecting from the set yeah that works in April would succeed every time but that's not you want to test disabilities and getting run everything lies and that's the way to test it that's how you know when something is working and it's not working we have not you presented but we also run everything life for for every single bunch so it is very important like many books we have fallen in and in open side and that are related to health services and were not tested life and we have marks for those sets so not is not tested and before closing of this this is a python conference are here 3 all libraries for for the integrations like come with these for sending messages by come with a library is actually used by salary and it supports transports and every transport is basically a messaging technology you could use you can use rabbit and human would be our eighties and some other sort of technology is that it supports idea and you have an so there really is a distributed task manager there was a presentation of online about it although ability to have distributed workers doing something based on messages and so itself uses RPC implicitly to tell workers why they have to and other messaging is an RPC library and that's what we use in OpenStack to send our PC messages through services and it also supports the will of the hazard architecture to support many brokers just supports Robert q and q bit for now and we're working on and your employ no for it and is the sum of all messaging technologies that you could use of you probably already know them like have come 20 different Q and a qubit family you have to keep the D which is the broker and it supports 0 . 10 and will have to support so I want we know as well you can have you could use to the probe which is fully and you on going no and to be displayed for routing messages throughout your system it was very much any questions thank you for coming microphone from view 1 graph with life but and then sort of you thought and I was very good the quite curious about how do you do your systems integration testing like do have some some
automated system integration testing of setting up a cluster services and so on
and what tools do so you have is that we use Gary for career you have it you something about the had this tool with the which is our testing integration tool that basically get certification from varied and it runs a Jenkins drop everything gets a budget and those Jenkins will his soul OpenStack completely in a single node and will tell us that we have life says that coal API and they would send messages to other cultures and then like simulator alive environment like spawning new would machines taking it down creating new orleans billion in bullion screen images and the latent images all that kind of things so it's because life we do have automated automated tools like Jenkins is basically the 1 that does everything and we use this fact to insult pretty much all the tools in those granting constructs by thank you for your work I have a question should do you didn't talk about security of if it if you run this messaging investor security short of so right now the institute is pretty much done by winding everything on your work in this later we have some work going on on signing messages and encrypted messages before sending them sending them through the whole of and applied so speak there was a talk about of Marconi they that was yesterday I will 1 of the things the good things about my money there was presented that it is good when the message broker is not good enough out 1 case is a special security like we have this agents running into the machines and we don't want those ghosts edges to talk to the central broker so much money would be good for that use case where you can just set up a new service that doesn't have to take a high load of messages in your infrastructure and you will you isolate everything for your message right so the security that how it's on up as the analysis just by binding of and our network and dozens and we'd all anyone to talk to that except for the services running in OpenStack deployment and I guess that we have some more going on for to assign messages and include messages before sending them through the White the right you have another question do you have a way to to make the dependencies between you uh services usable because someone I see this communication because it looks very clear and simple you just put a message from the boss and somebody else will get at that and that's just a way to with their father for the services to communicate with each with each other yeah and you can easily build a spaghetti dependencies systems by just using a very clear about this so all you prevent this politically like we don't have any any association between services that say like OK we can depend on each other sociological 1 good design decisions are being taken by we cannot make serious age-dependent is the answer is the dependent services so that try to figure out a way to do that which basically means creative services that has come for the lab but yeah it's a logical like cycle dependencies of in my opinion about but they're not always that like everything is so far but we try to avoid them as much as possible and sold on like logically would have everything explicit so since we know which service abandoned strollers logically speaking of functions wiser feature-wise we know that we cannot create cycles and some of the services or we try not to much and use the mikes sorry and that's explicitly written down somewhere in the code yet not somewhere in the kernel with this it is in the cold of labor have also documentation about it the region which pages and then the the condition of the church service and operations will others because you have to know how to the whole thing if there are no further questions that like to thank the speaker again thanks thanks for coming
Physikalisches System
Offene Menge
Systemprogrammierung
Softwaretest
Desintegration <Mathematik>
Code
Relativitätstheorie
Systemintegration
Wort <Informatik>
Computeranimation
Keller <Informatik>
Aggregatzustand
Distributionstheorie
Bit
Subtraktion
Desintegration <Mathematik>
Code
Computeranimation
Eins
Netzwerktopologie
Physikalisches System
Web Services
Vorzeichen <Mathematik>
Parametersystem
Graph
Datenhaltung
Reihe
Systemintegration
Red Hat
Systemaufruf
Physikalisches System
Quellcode
Integral
Lebesgue-Integral
Generizität
Dienst <Informatik>
Zustandsdichte
Menge
Rechter Winkel
Offene Menge
Datenverarbeitungssystem
Basisvektor
Mereologie
Strategisches Spiel
Authentifikation
Ordnung <Mathematik>
Web Services
Dienst <Informatik>
Web Services
Desintegration <Mathematik>
Bildschirmfenster
Bus <Informatik>
Bitrate
Computeranimation
Integral
Design by Contract
Web Services
Telekommunikation
Bit
Quader
Desintegration <Mathematik>
Atomarität <Informatik>
Systemaufruf
Term
Computeranimation
Integral
Dienst <Informatik>
Bus <Informatik>
Bus <Informatik>
Message-Passing
Telekommunikation
Stabilitätstheorie <Logik>
Subtraktion
Punkt
Netzwerkbetriebssystem
Atomarität <Informatik>
Hausdorff-Dimension
Kartesische Koordinaten
Computer
Term
Division
Computeranimation
Virtuelle Maschine
IntServ
Knotenmenge
Datensatz
Web Services
Perspektive
Grundraum
Zentrische Streckung
Befehl <Informatik>
Elektronische Publikation
Hardware
Sichtenkonzept
Computersicherheit
Datenhaltung
Reihe
Magnetooptischer Speicher
Schlussregel
Physikalisches System
Elektronische Publikation
Integral
Objekt <Kategorie>
Gefangenendilemma
Dienst <Informatik>
Menge
Rechter Winkel
Festspeicher
Server
Message-Passing
Logik höherer Stufe
Instantiierung
Aggregatzustand
Computeranimation
Router
Metropolitan area network
Message-Passing
Punkt
Transformation <Mathematik>
Desintegration <Mathematik>
Message-Passing
Computeranimation
Serviceorientierte Architektur
Zentralisator
Resultante
Offene Menge
Bit
Demo <Programm>
Punkt
Prozess <Physik>
Gemeinsamer Speicher
Formale Sprache
Familie <Mathematik>
Versionsverwaltung
NP-hartes Problem
Computer
Komplex <Algebra>
Skalarfeld
Raum-Zeit
Computeranimation
IntServ
Streaming <Kommunikationstechnik>
Web Services
RPC
Reverse Engineering
Bildschirmfenster
Mixed Reality
Stützpunkt <Mathematik>
Notepad-Computer
Einflussgröße
Serviceorientierte Architektur
Umwandlungsenthalpie
Parametersystem
Lineares Funktional
Zentrische Streckung
Typprüfung
Kategorie <Mathematik>
Datenhaltung
Reihe
Systemaufruf
Bitrate
Algorithmische Programmiersprache
Ereignishorizont
Lebesgue-Integral
Arithmetisches Mittel
Serviceorientierte Architektur
Generator <Informatik>
Dienst <Informatik>
Würfel
Rechter Winkel
Festspeicher
Server
Dateiformat
Information
Schlüsselverwaltung
Message-Passing
Instantiierung
Aggregatzustand
Telekommunikation
Subtraktion
Wellenpaket
Gewicht <Mathematik>
Quader
Mathematisierung
Transformation <Mathematik>
ROM <Informatik>
Mathematische Logik
Term
Code
Virtuelle Maschine
Datensatz
Knotenmenge
Datentyp
Spezifisches Volumen
Optimierung
Speicher <Informatik>
Bildgebendes Verfahren
Einfach zusammenhängender Raum
Beobachtungsstudie
Videospiel
Kreisfläche
Protokoll <Datenverarbeitungssystem>
Stochastische Abhängigkeit
Systemintegration
Datentransfer
Peer-to-Peer-Netz
Schlussregel
Physikalisches System
Binder <Informatik>
Ausgleichsrechnung
Integral
Design by Contract
Offene Menge
Hypermedia
Basisvektor
Partikelsystem
Computerarchitektur
Distributionstheorie
Bit
Gewichtete Summe
Formale Sprache
Familie <Mathematik>
Skalarfeld
Computeranimation
Softwaretest
Web Services
Minimum
Serviceorientierte Architektur
Web Services
Softwaretest
Lineares Funktional
Sichtenkonzept
Ähnlichkeitsgeometrie
Kontextbezogenes System
Lebesgue-Integral
Dienst <Informatik>
Druckverlauf
Datenfeld
Menge
Rechter Winkel
Server
Message-Passing
Subtraktion
Abgeschlossene Menge
Implementierung
Kombinatorische Gruppentheorie
Transportproblem
Term
Code
Task
Informationsmodellierung
Knotenmenge
Datenverwaltung
Programmbibliothek
Luenberger-Beobachter
Inverser Limes
Qubit
Programmiersprache
Videospiel
Graph
Systemintegration
Hasard <Digitaltechnik>
Physikalisches System
Quick-Sort
Design by Contract
Integral
Dreiecksfreier Graph
Mereologie
Programmierstil
Wort <Informatik>
Computerarchitektur
Lie-Gruppe
Telekommunikation
Code
Computeranimation
Homepage
Kernel <Informatik>
Virtuelle Maschine
Knotenmenge
Arbeit <Physik>
Web Services
Vorzeichen <Mathematik>
Diskrete Simulation
Tropfen
Bildgebendes Verfahren
Serviceorientierte Architektur
Touchscreen
Analysis
Softwaretest
Assoziativgesetz
Videospiel
Nichtlinearer Operator
Lineares Funktional
Konstruktor <Informatik>
Euler-Lagrange-Gleichung
Datennetz
Benutzerfreundlichkeit
Computersicherheit
Güte der Anpassung
Red Hat
Physikalisches System
Integral
Entscheidungstheorie
Dienst <Informatik>
Rechter Winkel
Benutzerschnittstellenverwaltungssystem
Last
Konditionszahl
Dreiecksfreier Graph
Programmierumgebung
Message-Passing

Metadaten

Formale Metadaten

Titel Systems Integration: The OpenStack success story
Serientitel EuroPython 2014
Teil 107
Anzahl der Teile 120
Autor Percoco, Flavio
Lizenz CC-Namensnennung 3.0 Unported:
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 10.5446/19971
Herausgeber EuroPython
Erscheinungsjahr 2014
Sprache Englisch
Produktionsort Berlin

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Flavio Percoco - Systems Integration: The OpenStack success story OpenStack is a huge, open-source cloud provider. One of the main tenets of OpenStack is the (Shared Nothing Architecture) to which all modules stick very closely. In order to do that, services within OpenStack have adopted different strategies to integrate themselves and share data without sacrificing performance nor moving away from SNA. This strategies are not applicable just to OpenStack but to any distributed system. Sharing data, regardless what that data is, is a must-have requirement of any successful cloud service. This talk will present some of the existing integration strategies that are applicable to cloud infrastructures and enterprise services. The talk will be based on the strategies that have helped OpenStack to be successful and most importantly, scalable. Details ====== Along the lines of what I've described in the abstract, the presentation will walk the audience through the state of the art of existing system integration solutions, the ones that have been adopted by OpenStack and the benefits of those solutions. At the end of the talk, a set of solutions under development, ideas and improvements to the existing ones will be presented. The presentation is oriented to distributed services, fault-tolerance and replica determinism. It's based on a software completely written in python and running successfully on several production environments. The presentation will be split in 3 main topics: Distributed System integration ----------------------------------- * What's it ? * Why is it essential for cloud infrastructures? * Existing methods and strategies OpenStack success story ---------------------------- * Which methods did OpenStack adopt? * How / Why do they work? * What else could be done? Coming Next --------------- * Some issues of existing solutions * What are we doing to improve that? * Other solutions coming up
Schlagwörter EuroPython Conference
EP 2014
EuroPython 2014

Ähnliche Filme

Loading...