Merken

Keynote: So, I have all these Docker containers, now what?

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
b and the world and theft
they enter on your you may think you can
extend the untimely applied you're talking about my last big party members was of Python Picon J. P in Japan last year I think
it because was really fun without of much of soldier Japanese and my Japanese is getting better but not a great my Spanish is really really bad the Spanish and Japanese are very similar so maybe I should learn both together and it never really seriously out are pages so that is the thing to make sense the region by container and having containers having lots of containers because
ultimately everything can be compared and we do have lots of containers we want to know what to do with uh and will not ask you some questions later and see how far you are along with moving towards containerization so basically when we have also contain is what we do then and this is a problem we
face a this is a data since this is a Google data center in Iowa USA that's a place called Council Bluffs uh and this is where the big data centers and then I leave out following triple-B Bill account with machines than basic cluster so classes are 1 of the conference we have internally but these classes are broken down into cells so small that cells of small and we have many cells the cluster and this will probably a sobering look at the base of about 10 thousand machines and that acquire knowledge and this is a huge amount of compute power plus a compute power we need to make it available to our engine is a software engineer we have developed so how do we go about making this compute power available to al undeveloped it
was something like this this is what a developer does both doesn't context 1 thing do more you
see there then we see that we don't
want injuries have select from rack select machine and say they little money on the machine under SSH except keepyour a binary of it's machine SSH inter-machine stand up my what processed
a server or whatever may be looking to many machines model times that that's not gonna be possible computerized machines fusions of engineers huge amounts of just about so how does it
happen the basically we have a configuration file and in this case is called a bold configuration of any any India recently and nobody there had a bold and if you're familiar with bowling start we get that the movies together talk about bold because Paramount Pictures on it and it's kind like 1 about was kept secret so we have this thing called Baldwin only a little little about it with time so the sum is really good to show this in the context of what little but later which is the ladies the call of the ball configuration file and what the developer does is he created jobs that case involved the police job Hello world so which cell he was running time by for the set of the cell a few thousand machines this case is saying is called I see some random cell language rights and the specified for binary trees in this case that how well what's there was 1 how world will want and this is gonna be 5 binary statically linked for this dependence is with within so effectively we can manage to much anywhere without having to worry about online princess and that includes web server well so it quite big for about 50 megabytes so he specifies the path to his binary all her binary and unfortunately happens to many males suffer engineers 1 of female suffrage inside that encouraged women to be useful for engineers and arguments better specifies monuments of binary past body environment in this case will specify what pulled to 1 this is parameterized then we have some requirements and some resources that it is important circle back to this estimate so we can specify I'm afraid and how much space and CPU and ultimately we say how many we want to run so in this case we would run 5 for replicas of this job 5 tasks effectively and why 5 1 0 did you scale 10 thousand this was rank families machines do so I'm using the reference and thousand opposite so once you finish this how where she types in a command with part in the config file and that gets pushed out from this perspective old schedule and what happens then is that if I were a period time in this case matching is 40 seconds 10 thousand tasks or 10 thousand instances of that job stop and it takes 2 minutes 40 seconds roughly we do phase the role that would result when children do well once uh 1 of the key factors here is the size of the binary 50 megabytes 10 times 10 thousand about 40 giga bps of life and will be caching of 1 quite a lot we had to move around between 10 thousand change that is of huge amount again but eventually get to a point where we have 10 thousand men or the they
would look quite and little by the 2nd and all the side if
this is not all to Google that's not assimilate but often we came up with a name because it public similarity eventually reasonable uh and bold lines with itself so we show how on old masters and configuration this case we have a bold master which is highly replicated we have 5 copies of that for resilience and we have a lot of other things the down here and machines 0 machines use on racks 1 of finger boldly we have a schedule we have some configuration files and binary what happens is the developers the engineer crater his or her binary uh and these are massively distributed and parallel build system called pharmaceutical basic study available now called basal so we make this open source that our own build system elevated 1 article base will be a z of the order American B a CEO the or the canadian BA Z in the against a confusing believe me get the countryside the to by Rowson roots so basically he or she creates a
binary which is out and and it gets stored in storage for cell and then they push the
configuration file configuration Vargas computable loss that we have a persistent Paxos by explore consensus-based and what happens then is is schedule look around comes along and says they what is the desired state we should have this 1 and 2 we had this 1 is she's 10 thousand atrocities as that in not running which have 10 thousand that make sure that happens fix that and so
it goes about planning running of the same thousand tasks and it creates a plan and it's not tenable master horrible must make decisions themselves in the Baltics on these machines to run this particular task so take communicated these task ultimately run inside a thin container right so it has a container around it is not just 1 in a binary it is containerized very light weight should contain a does not doctor is not science based the Holy ultimately reported binary over the from storage and will start
with the most heated most Hollywood whatever dates and now in many many people never learn a model copies of
and so that's what we had 10 thousand but if we look at it a little bit closer we find this 1993 1 look quite the same doesn't expected this is a highly available service we expect some the lesson of the number of tasks among over time due to the way we operate and that's interesting so let's look at that look in more detail so thank things
failed but failure is kind of more generic term here and there are many reasons for failures and 1 remain reasons for failures and particularly for low priority jobs is preemption and look at the top of which is production jobs we have very few various and most of the dancer machine shop we're actually schedules some maintenance on machine and we take machine down that task totalling and she wouldn't be rescheduled often across the uh we have a very small number of mentions down here production jobs which are things like mapreduce is batch jobs they get printed with there had to be printed fact the calculation generous says that for about 10 thousand tasks about 7 or 8 of them will be not run in any given time because of preemption could be about to get to some arose but it want 1 that particular time and we see other things here we see again the the optimal of the blue bar which is the
change of them to be much the same as production and we have some other things as well uh of resources for a small number of machines and further we have as many trust as many machines we have missing values were given now we expect that if we don't panic a machine that it is possible to 1 of business and another
interesting thing is how we try to make efficient use of our resources uh so we have CPU's we have memory we have described we have no and
sometimes it's quite possible for 1 task to be used using loss of of memory for little all by specialist CPU married memory it was machine uh then you may be wasting money and resources is
1 is resource standard and these available resources these white policy and this is the this example here is actually of those machines with the task of leisure machine direction containers please don't know as it could to compute engine these are all version machines these bars individual past and what we can see here is the somebody's machines have available capacity available RAM available CPU and we have a here we see but different situation where we have maybe some of available CPU and others with a portion of a program and vice versa this here and this here is called resource kind means we're not actually making use of resource so we have a spare memory capacity basically
capacity is being wasted effectively so what about challenges is like a Texas possible to try to stack these things in where will we get the best possible utilization of clusters say we will and that makes a match and make sure we have low CPU memory jobs finding that higher memory you jobs but also from multiple
tasks from the same that's extremely important like and then come back to this accumulated surely and another interesting thing is that which is going to be a huge challenge in future when it comes consecutive but it would be really important so we saw earlier that a developer she's specifies what resources she wants use what he wants use a hundred megabytes of RAM 180
hundred megabytes of disk will 1 to years and that will be his
blue line appear servings wanted and will matches is influence the 0 resources are requested by the jobs in reality there this idea and so we have always is wasted space we because it's been allocated effectively Ford ends when jobs but we can use it all we do is we effectively estimated based on the run patterns all of the common jobs how much that that is blue line here so this is our reservation this is how much reserves specifically for the jobs and or we can end
it is we use what space that will you reuse that space for very low priority jobs again batch jobs and not reduces things that we want run we want to finish eventually but we don't really care what
happens it could be like that of some kind a monthly report that no given set against well alright MapReduce across a huge amount of data that may be important some point we'll just needs to be done but we don't really care needs to be so what i said we can
reuse and we can run jobs and that's really important that's how we get maximum utilization part of 1 of our machines in that center
and so we windows accumulates now gradually uh when oscillations is that if you have your developers spending time thinking about the change of thinking in terms of machines and you probably doing it wrong because it's too love a level of abstraction that today maybe is finding the future this is not the case that we need people be thinking applications will have to worry about the infrastructure which around mean anybody of pattern was a serviceman how important is anyway you care about the structure you were right you work configuration file build binary wonderfully thank you we when it meant care about how you did this money for me to make sure it stays run and we get efficiency by sharing of resources and reclaiming unused allocations and containers the fact the fact we continue everything that allows us to make uh make use much more productive those everything we run Mandarin container achieving contains a week we estimate uh whenever you photos are important and showed off came along and contains became the big thing right exceeded then darker and that could be confused and so now 1 of the things we talk about what time know is we run containers and we are pretty good
when containers which is why we critically native if you're interested in more detail what just talked about all this a paper here you don't you no 1 have
received for and you of nest white paper on ball that's good the cells were directed to do so I guess it's a much much more detailed this work in terms of
a simple application and how we can do this externally that would contain through cumulated there is a very simple pattern uh generally when we could be so his PhD in milk well myself and impact client that many these plants and when has given many instances of classes to be some kind of events in the system that we have the ability to run many many concurrent uh I request before you go on scale these things on demand and we may want to scale skill that much until we get a point we had to do replicas and shot and then cashes scale as well but we'll keep it simple just keep you warm I still want cash and a few days of class instances front and
so this overlap containers so I mean you the food containers and the view that she's rule spun up a docker containers a of the sum of the same number the again last year with ASR is how many of you had containers both hands and you sample topic and most having to change the doctors the future all contains a future we got this thing called the open container projects doctor appointee made there made what they haven't respect to enroll in a get behind it we have this common specifications which could containers things like chorus with a rocket container progress fall in line we have a common format the container is going to great but as
those of you who are not really familiar contained a few slides very few slides on containers and designed using the concepts uh additional obesity things in the old days we have a machine maybe that's a that's not bedrooms or in the color on the server and the machine would run system it would have on the packages installed a provided libraries and openness itself what's top
government when application and how many of you have a situation where the running 1 application and all the other applications on machine failed because that 1 application that the user will receive use all of the random graphs and machine and so will the other applications that this may be a very low
priority at 1 you can really care about thinking that's really important this is never a good idea on a multiple applications along the chain because there's no oscillation between did whatever affects 1 application for the effect the other there's no namespace and they all have 1 view all of machine which to run that 1 view CPU 1 viewed memory when beautify our system they share library and so you can a situation where maybe 1 day you update version of a package it updates the library and 1 of the
applications is high and the running the model is not compatible with me their dependency hell and it is a Windows is still well how for
applications are highly coupled to the operating system this is a problem so we created
machines and what we did basically is stuck a layer on top of the hardware hyperbolas we now have an idealized piece of hardware which you could run multiple operating systems that we have this thin layer it looks like a piece hardware to be running but you machines and tag is a so isolation is that we can run applications in there and those who machine so each application is no isolated it 1 application crashes it doesn't
affect the other but since assuming efficient because we have this red big bottom here we have the offering system because you know when you sort machine you pretty much have to install the entire deviance directly this time central cycle inside window size and so
Casanova efficient all this story same tight coupling between the operating system and the applications and as anybody you try to manage lots and lots of those which aims to to provide isolation you know it's hard the new
ways containers and in this case we move up online remove above differences among provided the idealized operating system that the longer idealize
hardware they idealize operating system on which the command axis and independent library the so libraries here a part of the container to container has an application has with these dependencies it has its entire environment to become movies contain around anywhere we want to move from 1 machine to another from 1 1 time to another and from a laptop to the Dutch machine unintelligible that was to set top box all communication maybe mutual funds uh when we had doctrine and read ensures that right and this look at the
example uh so we have how applications PHP impassioned should be Python patches sorry and I do apologize that size Python so where we see page fatalistic replied from that I will change version of slides so which I think provocative and a Python wouldn't legs this problem to that page they're gonna OK so we have container so that we wanna run these components of application Python and Apache memcached let's go well obviously from Boston bottle or you things you could potentially use then cash my skill and let's go has its own libraries it doesn't have any common libraries with the other side we despite those libraries with the container in which last around and and then passed the and PHP or python 300 kids in a Python Apache Python whatever call and federal dependencies but I also have been Solingen chain-dependent as well but some commonality so when we actually pretty image we you share some stuff between them but also shared 1 time from the credit container they will
have to run dependencies packages revenue in the container None of these that we have a
server and again this could be a bunch machine it could be a laptop it could be a bimodal so that they could be anything much and we had the actual hardware and all of this should be maintained by a document in for doc is the thing that runs the middle by containers mostly synonymous with talking about a you're out of the container format and hopefully get comply with a standard and that's the vulnerable heading toward the doctor effectively controls the creation of these containers and the
management of the container so the end of it we would have a Python flask and angular memcached Westville contains so why contained so that
many important reasons for having containers but you can see just by looking at what we do know that that's anyway we can do it before doing away this is a perfect solution because skeletal we won't be so suppose that the smaller scales well uh why because is much more performance is much more before in terms of the fact that we don't have to do with their insulation they are all pretty much like the 1 on bare-metal said before must be much the same as much watching much to to get up and running which means you can chop allow quicker you can plot raised regarding the primitive increase repeatability so the whole problem where we have the development QA test production but we won't have repeated environments but we have a situation where when we testament Q and running project if I wasn't probit western Q I had a few of that situation is on like behavior hands remember those days right so the ball contains give us is ability to have a consistent environment the environment package container a so basically when we're running QA running part is exactly the same exactly the same environment so that's 1 of the great use cases also contains today but much more is the possibility of it but also about the 2nd Quality of service we can now do resource isolation as well using things like the groups in its namespaces weakness isolated the resources we can say we anyone is to have a a
bus ran under megabytes of disk will put 1 CPU
and ultimately accounting these things are easier to manage 3 theater Traceries years old a small composable units that can be tracked for usually they ultimately plausibility you can move these things around from 1 cloud providers for another uh images specifically the cop despicable running comparing that Buchanan easily run the same container in the different travel wider with them and machine a laptop and you can move from 1 machine to another as short as the shape of your class that you have a cluster machines changes you can move them around to be more efficient as something about so we have a full the efficient allocation of resources we can do that we have contained they
also really this is a fundamentally different way of managing building applications so I don't know that he's denied the
deciding my mistake this containers pocket and I that for you that is very easy to find a tutorial docking government this talk about the
selected nations that how many of you have heard bases how many of you can say this a hard work together and the Bobby these review great discrete words maybe he wants help pronouncement or be outside and the group with of this so that
I can that we provide assistance and that will be maybe saying around maybe I would say it rewards and some of them happy to be corrected for the ladies so that and we give new
insights into what we do a good reward so that should provide a context for a wide to native is necessary something we often miss out told you that we don't we provide that kind context from 19 the introduction of all that will be provided for you so debate is a big word
these helmsman would governor of summaries and so on false negatives government to continue to the next and this effectively orchestrator or schedule for the topic containers ultimately for other forms a containers of or already using it to schedule orchestrate rocket containers and it supports multiple cloud environments so mesosphere comes to get the where IBM Microsoft involved human kilometres away and isn't even when pretty much anywhere the command on your laptop vagrant antecedents crater full machine just such machines with a good thing up you have
accumulated cost and ultimately eventually uh we may have a situation where we can run through this across multiple uh Campbell writers it might be difficult it might be possible maybe 1 day you have your face of change the running Google and Amazon and much of the year as well as possible and emotional it can have so this is kind of inspired informed by everything we saw in the previously reviewed people and is based on our experiences but it's also go
like many department nowadays but the was the pattern of going on a Python based Java developer I since 15 years of a job the living standard to Google and wrote a lot of job codes in our i or it's like java programmers
anonymous friend since it's been foliations alright my last line geometry than alright in Python alright going the right angular right script and with more more interesting the job is getting better job is because the board and ultimately we once we had to talk about managing applications of machine decidable we about and some very quick
concepts of operations you which show you the icons so that when you see them you know in container apology service for the label replication control and node on the key concepts immediately sold stack the figure you like to determine origin so that the grains and sigh like and I think it's really hard to get around nothing when dangers about an abstraction is that you come you get too far away from the terms of a million people my cities of millions of people or service the idea of a replication controller at the node labels that container the point is for the most difficult 1 to get a very serious about Paul and
about nature and clusters so we have
a cluster put the match back to what we talked about earlier with bold well we have a master the master has a schedule Natalie PI the price of a it can use these talk tonight the nodes rewarding of encoded cupola and they have these things called pods when containers and report
shortly a proxy by which we can expose our running containers to the outside world
we have many names the class that is an abstraction so test that could be different depending which cloud providers will use and also more you won't have is the fabric of machines was like a flat shape on which we can then continue you don't care about it is cadastral joined together as 1 big flat space in which reminds us of Miletus thing is scheduled to take care of 1 and stuff for ultimately
this is basically the options for the clusters are laptop multi-node clusters are hosted or even managed on-premise or cloud based using but machines that will bear changing many many options is a matrix and here show you wouldn't stratified and this link will give you matrix of how we run key bases on what you want and chorus members we have
different ways of doing network in the network is quite tough and it will contain cable computing G makes it easy because of IP address and that often we
have to put of layering final so that should provide the ability to give an IP address people submit to a running machine or unemployed said this about politics
and the view from here consider the public this was from the so in the diagram here we have appalled we a container is the container server is contain and it has a volume by Dr. contains have volumes them at
different but very similar and so we have we wanna run this web server and the concert we use within To date it is to create a single
appalled the psychological height like if you wanted to run Apache and some of something else alongside it you'd running on the host machine that's the same as the pot so anything you would want to give us a machine run in the park these are the atomic units Saskatchewan accumulated this simple cubic scheduled we talked about jobs early immutable good-natured schedules pop your
contained is 1 of inside the pores the thing wrapper that these are ephemeral these alike is energy that it would ease this Texas cattle analogy that only like it from the vegetarian the crops silence and and Pontil like crops don't care about and we haven't we you
don't care about your individual and plants growing we have flowers for the given name water and you talk to them as well as the care about them but you don't care about your crops they say consular crops they can come guiding we replace all absolutely the same idea can take 1 place to another
and ultimately to make things simple now he don't have to worry about a party wanna run a single container you save bonus contains only it will create a plot for you and you still have to think in terms of point media monitoring but you don't have to create a point in the same manner containerful made it will create a portfolio that said pods are the abstraction that would get around a little bit more information about imagine a
scenario where you want to have something as simple as we get up so maybe a post to deploy touching area where we have a will developers to knowledge and to get you want the status of made % abduction or may understating service so you have a thing called get synchronized and is talking to get monitoring your would want you project and get it falls and the changes and it licenses somewhere on the disk and you server can insert the content this is to tied together
they they work together and make sense for them to run side by side when 1 goes away never goes away so we can run in both the same point so now we saying this
logical host part this 1 2 containers and his case gets synchronize them and node j or classes and we have a shared volume considerable interest eventual is a tightly coupled together so when 1 when applied dietary dice together is making sense sentiments separately in my in your the way you architect things but it doesn't have to they share
networks based on input space they had the same concept of localized and they are the ephemeral and I think in terms of things you would run on a single machine
so volume was someone in when we talk about audience that there are 4 important remember things which you could really so the volume is basically bound support encloses it happened and this is something where we can like date for read data from I think we have many options when it comes to the volume Dr. already has volumes and this is slightly
different similar so to the container when the pulled the volume looks like a . 3 and what thereby by buying such like and what amounted is determined by the volume so the 1st 5 we had is an empty directory so whenever trade support it creates space somewhere on disk on the local disk and taken basically share that volume between the lives and dies and the plot the only exist what part is is there so it could be you'll get summarizes license stuff to this fully in the red by the Apache Server or whatever so and yeah care when the pump way that space because ways scratch stages temporal data there's
nothing stored it is important to you I can even be but on the results that could be and have of and strength the efficient much faster as well say that's what computers attached to the
4 you get for us as well ethnicity foolishly yes to specify what time it is so empty there is 1 of the options this 1 is highest path where we can actually map part of our system of the node to which the point is running into the part so this volume is actually effectively a snapshot of most natural and linking to the file system will get from 1 in machine test useful 3 configuration dates and stuff is also kind of dangerous of all because you may be that the state from the node may change in such a way that when you everyone upon 1 missing to another view money under the scheduler once the pod when given the same may see different view of what happens so it no longer becomes uh completely isolated physical dangers things the might work for you no 1 is in a fast and other similar services like just north of Gustavus fasciculus conversely anything was you know uniform and there again and effects become mount effect policy
on part the space into a contains this doctrine all sees
a cloud provider persistent storage the system block storage that we call position storage in Google Anderson called on the latter about storage tank I think this is persistent in the ACD they come right from the from the data from the disk and you will always be there were the plot goes away or whatever so what we would like to use cases creates the volume of 1 minute traffic light at the corner disk in the cloud provider which stores data and mounted it once upon the couple goes away the data store what comes along locomotive and also with Google cloud platform catch anatomy and
multiple points of so some patterns for parts of
the first one is the site of action and the Bayesian logic places like and I go I guess in this case the is application of Python and there is the then-girlfriend about that
IJS applications a bicycle is a did synchronized with the slides going from a to a central ambassador in this case something that acts on behalf of the actual 1 container this is a 2nd container a it proxy effectively allows the PHP applications may cause and and had reduced particle shots so we can make will have 1 service that the PHP application calls reads writes and it policy can do all the hard work deciding whether to read from a master all right the region of this label right from most unavoidable about pattern where In this case we have right this 1 in uh and we want to monetary rewards 1 for the plots but we need a common format for monitoring so in this case we actually that the output from the latest monitoring using that contain it up to contain will be plugged into someone
should so that kind of adapt ball happening within the thing that is a kind of examples of where it makes sense to have appalled sometimes that makes ends up being just that
if you afterwards but pods make sense to you so labels labels basically the single group a mechanism of incubators and this allows us to group things would be to build applications like attached so never prevalent import give label labels are key value pairs so this case titles completely arbitrary method somebody's going to meaningful to Cuba latest mostly could be anything meaningful here so we've put labels on pods and we can say the reduction of the publication of user data like to say give
me a point that is and by the so you would say something and so
we have if labels to devote to this has we have a version 2 . have different actual applications 1 from that makes a lot more has pointed out of many labels I surprised myself my slides into place well it makes more sense of replication controllers because replication controllers all things actually manage the monopods number and I said before that we created 10 thousand tasks and we push them out to this is the origin of all lost this digital comes along says that they should be run in that and not a fixed but this is the same thing replication control is responsible for managing your desired state you say this is the way I want to be want to have make somebody's pods basin is contained in the template like somebody's pods basin is contained place I want you to maintain that stable lies job at replication control so basically what they do is they work on a constituency of a label type of labeled in this case the ocean was the 1 is what they select this application controllers responsible for all points with label but was the 1 and we turn it on add to it so it's job is to make sure it was to In this case we also have a number of occasions trauma that has to be to appalled person with feature or anywhere any size symmetry always 1 religion a cutaway what
is that this is kind like a controlled said repetitions 1 be controlled simply that it says that it desired state emission honey with the 1 in which a foreign the foreign the foreign below
free when has a good sort of FOR and therefore the firing that that's the that's the 2nd way so it is continuously monitors the state of nature we had 1 run it also much the 10 so we provide
template which is the pumpkin plate which contains a container image definition and we want to run the past and replication Angela it doesn't fade appalled but when we create a replication control in this area to these politics as now there's also learning should start so but I was a middle plugging replication controllers after we apart from the states go managing contains and finally
we the services and services of how we actually expose our 1 and stuff uh and we do this through this service here which plays a virtual IP addresses which is has a constituency all of polled
based on labels for it again whenever labels on here which only 5 so basically certain parts of certain label on a constituency of the service and when requests coming from
clients you will a balancing frost the the bottom part of of the Goddess movement around so they could be 10 thousand nodes and 10 thousand ballots on different nodes and he would like balanced across the running time and memory and he was brought down problems and eventually it will have much more intelligent support for mobile this is it's useful exposing internal services within communities the face bones uh mons services to clients the study which was the shortly it is not advisable try pages but also a DNS name for it into service discovery and
we'll move on to this Canary example so you understand the concept of an area with a furious eventually when you have a situation where you have a lot application you wanna try a new version of you may you have 1 instance to instances of that when application of different as
a traffic will be possible this new versions of some organs in your questions you can do a b testing against and to make sure that the new service works it doesn't about back the stars you can push the change for this is a similar situation we have a
virtually was the 1 mostly closely to the replication controllers and pods the service the cares about his labels typographic and so the service has its consistency all 3 of these pods but these management of replication control so that's how I works but shopping address exposures that's clients 7
Montague bases it was kind of ideas we have plots promoters symbols Blair is why it's important this apology body service and we have a and then Memcached the program apart services and replication controller of the service should develop from the how
it looks to the developer
group so this is how it looks they specify name they can specify the image this is a documents now it could be a different image format in the future for different topic contained forward all and the fact and
that is deliberately to something at the at page because but the thinking is that the virus resources
128 maybe they needed it said that is that if I am a CPA she the poaching and unlike your slices of CPU and nobody getting there but it's like 500 bits of CPU and accumulated began to read many of with of was when the intensive heart brother had like percentage but that doesn't work because have you recall and have powerful the that's how we specify the view the poll protocol TCP Americans on or maybe and that again we cover the case as well that's how once been replication and all this other configuration files as well as
services offered to and scheduling a moment we saw the complexity is hidden in a Google is this simplification makes this turning is based on point selection so we won't have the points within the basin selective and is based on like that so uh how much capacity this and now we have is a cable were part for me if I had multiple nodes a command iPod uh under on the 1 that has the least resources consumed by running parts and that's a priority In the future we have resource-aware scheduling so we can do kind of what we did we do back and go will be try to make maximum utilization of CPU memory good radius is 1 . 0 has also this week and notes from the 1st of time that was called important there has been a source for every
and we have a product called dual contingent uh which I will talk by actually not so much but is a good way of 1 the there was of pitch hence accumulate is under 2 more contained and usually and a roadmap for conditions is there is a kind of sparse the moment because you come from 1 . 0 that are not assigned in a way that for the next release is the 1 . 1 and 1 on the right and current auto-scaling and the ability to alter scalable nodes fundamentally based on you know what you we
are considering gene is a major speculators and it manages up time
for you you have to worry about the master his case will take care to master avoid Università mostly call connection was the 1 the problems we have mind commissions is high availability so we don't have that replicated master said with syllable so there were 2 days have multiple clusters to do high availability but we look off your masterful you make sure it's running and and you don't worry about it we will make sure audio clusters available for making sure you must away so we can resize uh these Finkel is going to again centralized looking and we can prove look 1 place in the dual developed council and it also supports the PN so you can actually have your points uh inside your network and provenance the very
quickly and we had to change the set of earlier to make this work this is a cluster we have q consul get notes so we have to know during these machines in a cluster and I can look at him here this is a good developers
conference the involvement of the smaller we're going to be an instance is here can
see by running the genes that accommodation well they soon middle are the notes from uh I had a single instance group
which has 2 instances and this is the thing that manages the size of a cluster and blow here we have contended clusters missing we have 1
cluster OK there we go to the this
I've got very little screen real-estate across here going on we go here which
representational was 1 currently so we have these appalled that is a appalled this is a service prices space external and internal this is a service and it is not a service of my skills 1 in which the role playing but of when it the other thing I can say we have a front-end service we have non-cash service and we have left over to the Portland from of the apart so
this sort of called frequently and
the that's why I don't want and they would just go to 1 . 0 to my whom I went on and break the should be I had a 1 and a half century demolishing because there we have display then we have applied in the vault and myself idea was spun up my sculpture quickly but they haven't this is what is 1 of PHP is 140 . php currently but I'd be going trying to get times taken completely but problems of
fossil angular so anybody else happened fast likely I should talk to your basically all my my batches system I my Python skills of reached this 3 styles of abuse the told you go about doing this there 2 cultural 1 the great and integrated controller we have
a father already created and we can create that and now we have pods the replication during that is where the smaller so
we have some fun templates and the front-end Aussie control Signed Nixon wanted is actually different application because my
windows all screwed up much trouble we have
it running so this is the IP address of the services we can see
here and this is the
application running this default the rebound to double go against the will to Europe right to my children to fix before and so we have an update and be rolled out they are not really easily to cluster do this for an update from after and I'm going to play that down
from deceit visualization
and I will go always 1st mention and I'm going to update features of front control those was to have another place in the control that's changes . 1 by 1 to roll out new version that we have 3 . 2 2 . 0 and C 1 . 0 so we get a 1 a 1 . 0 is that we have a 1 . 0 2 then integrating in the 2 . 0 port I'm going to give it a bit of a 1 . and eventually you we only have to use when you're parts and we get rid of the 1 . 0 control of you need anymore and the get back to no that is why refreshed which
is not only few 11 words movements 1 problem then can again there what's believe that government
as well I should mention that the public interest in that time what is the
commander us even more I have
ways to get it to complete skeleton summary the 2 on the status of 6 because the fiber because the about 12 is the number of
replicas is the stay in in and we have 5 unemployed the position to that now we have 5 minutes behind also the developers control that is to wrap up on the whole thing discrete talk about the last bits and pieces that's how we
visualize this we visualize it
using the API and a proxy thank causal support approxi we just
pointers and Jason D. Jason jobs could be just there's all JS plant three-dimensional views suggests they intend to
contain original plus state and we have a single managing since group and that one's word on nodes and nodes 1 of n-dimensions group and have a single this is the manager that crates and is responsible making sure running so that's actually from the cluster of nodes and we have a 10 played by
which we can try to create new nodes on demand so we can resize that that that managers has been very easily uh and yeah this that for just we can also
create clusters using tools such as the Google developers Council uh it would appoint manager and careful that was to give an example this very basic careful will try to costs avoided the 1 allow you to resize it don't resize right replace it completely which is really what they can create classes the various somewhere
and our civilization some frequently asked questions are answered and documentation that I could
spend entire hours and will be subject so if you have questions will be outside all day on the Hilbert promising and who made is open source so we want your help make better said please contribute to the Committee's yeah questions gates IRC receipt free of and that 1 has to contain this it's a very popular place and so on Twitter activities or I think 3 questions for me but also the now we can find number
and what we said feature
genetic if we have time for 1 of the questions at the beginning and you were talking about the bark and like 5 masters that you Our running and does figures are based on the data center or part of conversation cell itself and so we break up into only showing
belongs the that's where my limited knowledge of health the complexes of helper words Bobina squealing sorry yeah that's how well that's why we have most of your reading do you find most of all classes from all the cells infected this is the talk very interesting when you mention that when you compared the enzyme and containers and he even used the user of IBM house access is very difficult to escape from the
upper right so etc. Our how these UT in the current and contains implementations this work-in-progress size this is about security of containers on your
reading becomes too much of it is getting better with time uh initially we had problems with the article we have all this is called such that we made patterns the open for interesting but is getting better ultimately doctrines of becoming more secure with time uh ultimately do multi-tenant maybe currently have multiple customers applications 1 side by side may not be the best idea but we have to tackle that ultimately we have to make sure people more comfortable that they can run or their jobs on containers securely under required a EPU working and that's ontology occur that question is for personal that that that that sort
of thing and that's the 1 can find the outside comes components of a webpage fields Python so why it
so it defines value Monday to go on different operation thank you as to this and but 1 of our of the virtual and labor market thank you haven't fj
Roboter
Metropolitan area network
Uniforme Struktur
Red Hat
Vorlesung/Konferenz
Baum <Mathematik>
Computeranimation
Homepage
Rechenzentrum
Mooresches Gesetz
Virtuelle Maschine
Klasse <Mathematik>
Zellularer Automat
Vorlesung/Konferenz
Software Engineering
Leistung <Physik>
Gewichtete Summe
Punkt
Formale Sprache
Familie <Mathematik>
Binärcode
Raum-Zeit
Computeranimation
Gotcha <Informatik>
Gewöhnliche Differentialgleichung
Prozess <Informatik>
Prozessfähigkeit <Qualitätsmanagement>
Trennschärfe <Statistik>
Randomisierung
Vorlesung/Konferenz
Zentrische Streckung
Parametersystem
Prozess <Informatik>
Systemaufruf
Ausnahmebehandlung
Matching
Kontextbezogenes System
Binärbaum
Frequenz
Teilbarkeit
Endlicher Graph
Scheduling
Menge
Rechter Winkel
Zellularer Automat
Server
Programmierumgebung
Instantiierung
Mathematisierung
Zellularer Automat
Zentraleinheit
Task
Virtuelle Maschine
Informationsmodellierung
Benutzerbeteiligung
Rangstatistik
Perspektive
Datentyp
Softwareentwickler
Konfigurationsraum
Schätzwert
Videospiel
Kreisfläche
Zwei
Eingebettetes System
Elektronische Publikation
Mereologie
Beobachtungsstudie
Scheduling
Virtuelle Maschine
Open Source
Ähnlichkeitsgeometrie
Physikalisches System
Wurzel <Mathematik>
Elektronische Publikation
Softwareentwickler
Ordnung <Mathematik>
Konfigurationsraum
Gerade
Computeranimation
Einfügungsdämpfung
Gewicht <Mathematik>
Automatische Handlungsplanung
Zellularer Automat
Computerunterstütztes Verfahren
Elektronische Publikation
Computeranimation
Entscheidungstheorie
Task
Scheduling
Virtuelle Maschine
Rechter Winkel
Speicher <Informatik>
Konfigurationsraum
Aggregatzustand
Softwareentwickler
Bit
Minimierung
Zahlenbereich
Rechnen
Biprodukt
Computeranimation
Softwarewartung
Task
Metropolitan area network
Virtuelle Maschine
Generizität
Scheduling
Informationsmodellierung
Dienst <Informatik>
Bit
Prozess <Informatik>
Vorlesung/Konferenz
Stapelverarbeitung
Einfügungsdämpfung
Mathematisierung
Versionsverwaltung
Kanalkapazität
Zahlenbereich
Technische Informatik
Biprodukt
Zentraleinheit
Computeranimation
Richtung
Task
Metropolitan area network
Virtuelle Maschine
Festspeicher
Vorlesung/Konferenz
Emulator
Optimierung
Standardabweichung
Task
Task
Prozess <Informatik>
Festspeicher
Softwarewerkzeug
Kanalkapazität
Vorlesung/Konferenz
Softwareentwickler
Zentraleinheit
Computeranimation
Prozess <Informatik>
Mini-Disc
Mustersprache
Server
Vorlesung/Konferenz
Stapelverarbeitung
Raum-Zeit
Gerade
Computeranimation
Betriebsmittelverwaltung
Punkt
Extrempunkt
Mathematisierung
Kartesische Koordinaten
Term
Computeranimation
Übergang
Virtuelle Maschine
Nominalskaliertes Merkmal
Prozess <Informatik>
Digitale Photographie
Mustersprache
Bildschirmfenster
Vorlesung/Konferenz
Softwareentwickler
Datenstruktur
Konfigurationsraum
Tabusuche
Benutzerfreundlichkeit
Abstraktionsebene
Gebäude <Mathematik>
Softwarewerkzeug
Elektronische Publikation
Arithmetisches Mittel
Menge
Rechter Winkel
Mereologie
Pendelschwingung
Verkehrsinformation
Zentrische Streckung
Punkt
Klasse <Mathematik>
Zellularer Automat
Kartesische Koordinaten
Physikalisches System
Term
Ereignishorizont
Computeranimation
Client
Mustersprache
Vorlesung/Konferenz
Instantiierung
Umwandlungsenthalpie
Addition
Sichtenkonzept
Gewichtete Summe
Zahlenbereich
Schlussregel
Physikalisches System
Computeranimation
Rechenschieber
Virtuelle Maschine
Arithmetische Folge
Offene Menge
Stichprobenumfang
Server
Programmbibliothek
Dateiformat
Projektive Ebene
Kantenfärbung
Gerade
Soundverarbeitung
Virtuelle Maschine
Namensraum
Sichtenkonzept
Verkettung <Informatik>
Festspeicher
Zufallsgraph
Versionsverwaltung
Programmbibliothek
Kartesische Koordinaten
Physikalisches System
Pendelschwingung
Computeranimation
Zentralisator
Hardware
Machsches Prinzip
Systemzusammenbruch
Kartesische Koordinaten
Ideal <Mathematik>
Physikalisches System
Computeranimation
Endlicher Graph
Virtuelle Maschine
Informationsmodellierung
Multiplikation
Hyperbel
Netzbetriebssystem
Dreiecksfreier Graph
Bildschirmfenster
Minimum
Vorlesung/Konferenz
Set-Top-Box
Router
Telekommunikation
Subtraktion
Hardware
Stochastische Abhängigkeit
Kartesische Koordinaten
Ideal <Mathematik>
Physikalisches System
Computeranimation
Virtuelle Maschine
Datenmanagement
Rechter Winkel
Netzbetriebssystem
Notebook-Computer
Mereologie
Programmbibliothek
Vorlesung/Konferenz
Programmierumgebung
Mathematisierung
Versionsverwaltung
Systemaufruf
Kartesische Koordinaten
Endlicher Graph
Homepage
Computeranimation
Rechenschieber
Patch <Software>
Programmbibliothek
Vorlesung/Konferenz
Zusammenhängender Graph
Bildgebendes Verfahren
Softwaretest
Zentrische Streckung
Namensraum
Hardware
Gruppenkeim
Biprodukt
Term
Computeranimation
Virtuelle Maschine
Fundamentalsatz der Algebra
Datenmanagement
Softwareschwachstelle
Notebook-Computer
Mereologie
Server
Dateiformat
Vorlesung/Konferenz
Projektive Ebene
Primitive <Informatik>
Softwareentwickler
Dienstgüte
Programmierumgebung
Standardabweichung
Betriebsmittelverwaltung
Virtuelle Maschine
Shape <Informatik>
Einheit <Mathematik>
Notebook-Computer
Mini-Disc
Klasse <Mathematik>
Mathematisierung
Bus <Informatik>
Vorlesung/Konferenz
Zentraleinheit
Cloud Computing
Bildgebendes Verfahren
Computeranimation
Gebäude <Mathematik>
Notepad-Computer
Stützpunkt <Mathematik>
Wort <Informatik>
Kartesische Koordinaten
Diskrete Gruppe
Hilfesystem
Computeranimation
Demo <Programm>
Managementinformationssystem
Offene Menge
Kontextbezogenes System
Computeranimation
Scheduling
Virtuelle Maschine
Metropolitan area network
Multiplikation
Bildschirmmaske
Negative Zahl
Notebook-Computer
Wort <Informatik>
Programmierumgebung
Streuungsdiagramm
Offene Menge
Programmiergerät
Prozess <Informatik>
Mustersprache
Applet
Mathematisierung
Codierung
Softwareentwickler
Computeranimation
Standardabweichung
Nichtlinearer Operator
Punkt
Abstraktionsebene
sinc-Funktion
Kartesische Koordinaten
Term
Räumliche Anordnung
Whiteboard
Computeranimation
Virtuelle Maschine
Dienst <Informatik>
Knotenmenge
Prozess <Informatik>
Rechter Winkel
Datenreplikation
Gamecontroller
Skript <Programm>
Figurierte Zahl
Gerade
Softwaretest
Virtuelle Maschine
Proxy Server
Scheduling
Shape <Informatik>
Knotenmenge
Matching <Graphentheorie>
Minkowski-Raum
Abstraktionsebene
Natürliche Zahl
Klasse <Mathematik>
Service provider
Verkehrsinformation
Computeranimation
Matrizenrechnung
Multiplikation
Datennetz
Computerunterstütztes Verfahren
Binder <Informatik>
Netzadresse
Computeranimation
Konfiguration <Informatik>
Virtuelle Maschine
Virtuelle Realität
Stützpunkt <Mathematik>
Pay-TV
Notebook-Computer
FLOPS <Informatik>
Streuungsdiagramm
Metropolitan area network
Scheduling
Virtuelle Maschine
Benutzerbeteiligung
Diagramm
Einheit <Mathematik>
Sichtenkonzept
Prozess <Informatik>
Server
Ordinalzahl
Spezifisches Volumen
Computeranimation
Energiedichte
Bit
Punkt
Abstraktionsebene
Wasserdampftafel
Hypermedia
Wrapper <Programmierung>
Plot <Graphische Darstellung>
Term
Computeranimation
Punkt
VIC 20
Mathematisierung
Klasse <Mathematik>
Dienst <Informatik>
Mathematische Logik
Computeranimation
Metropolitan area network
Dienst <Informatik>
Flächeninhalt
Mini-Disc
Mereologie
Inhalt <Mathematik>
Spezifisches Volumen
Softwareentwickler
Datennetz
Gemeinsamer Speicher
Temporale Logik
Plot <Graphische Darstellung>
Ein-Ausgabe
Term
Raum-Zeit
Computeranimation
Konfiguration <Informatik>
Virtuelle Maschine
Mini-Disc
Mereologie
Server
Spezifisches Volumen
Verzeichnisdienst
Resultante
Soundverarbeitung
Softwaretest
Sichtenkonzept
Punkt
Physikalismus
Physikalisches System
Computerunterstütztes Verfahren
Computeranimation
Konfiguration <Informatik>
Metropolitan area network
Scheduling
Virtuelle Maschine
Dienst <Informatik>
Knotenmenge
Mereologie
Dateiverwaltung
Spezifisches Volumen
Konfigurationsraum
Aggregatzustand
Service provider
Ortsoperator
Mini-Disc
Mereologie
Speicher <Informatik>
Plot <Graphische Darstellung>
Spezifisches Volumen
p-Block
Speicher <Informatik>
Cloud Computing
Raum-Zeit
Computeranimation
Web Site
Punkt
Gruppenoperation
Kartesische Koordinaten
Plot <Graphische Darstellung>
Mathematische Logik
Computeranimation
Rechenschieber
Metropolitan area network
Dienst <Informatik>
Rechter Winkel
Mereologie
Mustersprache
Dateiformat
Partikelsystem
Große Vereinheitlichung
Funktion <Mathematik>
Kraftfahrzeugmechatroniker
Punkt
Anpassung <Mathematik>
Gruppenkeim
Kartesische Koordinaten
Schlüsselverwaltung
Ordnungsreduktion
Computeranimation
Abstimmung <Frequenz>
Stabilitätstheorie <Logik>
Punkt
Template
Versionsverwaltung
Wiederkehrender Zustand
Zahlenbereich
Kartesische Koordinaten
Computeranimation
Endlicher Graph
Task
Rechenschieber
Symmetrie
Prozess <Informatik>
Endogene Variable
Datenreplikation
Datentyp
Gamecontroller
Lie-Gruppe
Aggregatzustand
Software Development Kit
Abstimmung <Frequenz>
Virtualisierung
Natürliche Zahl
Template
Menge
Netzadresse
Computeranimation
Metropolitan area network
Dienst <Informatik>
Flächeninhalt
Datenreplikation
Gamecontroller
Bildgebendes Verfahren
Aggregatzustand
Beobachtungsstudie
Client
Subtraktion
Knotenmenge
Dienst <Informatik>
Abstimmung <Frequenz>
Festspeicher
Direkte numerische Simulation
Mereologie
Menge
Computeranimation
Homepage
Softwaretest
Selbst organisierendes System
Adressraum
Mathematisierung
Versionsverwaltung
Kartesische Koordinaten
Computeranimation
Client
Dienst <Informatik>
Datenmanagement
Flächeninhalt
Datenreplikation
Gamecontroller
Widerspruchsfreiheit
Instantiierung
Subtraktion
Division
Gruppenkeim
Plot <Graphische Darstellung>
Symboltabelle
Extrempunkt
Computeranimation
Metropolitan area network
Dienst <Informatik>
Verschlingung
Datenreplikation
Gamecontroller
Stützpunkt <Mathematik>
Dateiformat
Vorlesung/Konferenz
Optimierung
Softwareentwickler
Bildgebendes Verfahren
Computervirus
Bit
Mereologie
Sichtenkonzept
Protokoll <Datenverarbeitungssystem>
Division
Familie <Mathematik>
Zentraleinheit
Elektronische Publikation
Multitasking
Computeranimation
Metropolitan area network
BORIS <Programm>
Datenreplikation
Vorlesung/Konferenz
Konfigurationsraum
Baum <Mathematik>
Radius
Punkt
Momentenproblem
Extrempunkt
Güte der Anpassung
Softwarewerkzeug
Kanalkapazität
Strömungsrichtung
Schwach besetzte Matrix
Quellcode
Biprodukt
Zentraleinheit
Komplex <Algebra>
Computeranimation
Scheduling
Dienst <Informatik>
Knotenmenge
Rechter Winkel
Konditionszahl
Trennschärfe <Statistik>
Festspeicher
Mereologie
Einfach zusammenhängender Raum
Punkt
Datennetz
Betafunktion
Hochverfügbarkeit
Vorlesung/Konferenz
Computeranimation
Inklusion <Mathematik>
Gruppenkeim
Einfache Genauigkeit
IRIS-T
Speicherbereichsnetzwerk
Computeranimation
Endlicher Graph
Virtuelle Maschine
Menge
Softwareentwickler
Chi-Quadrat-Verteilung
Große Vereinheitlichung
Instantiierung
Demo <Programm>
Dienst <Informatik>
Bitfehlerhäufigkeit
Selbstrepräsentation
Raum-Zeit
Computeranimation
Gammafunktion
Touchscreen
Metropolitan area network
Datensichtgerät
Total <Mathematik>
Quick-Sort
Computeranimation
Inklusion <Mathematik>
Template
Kartesische Koordinaten
Oval
Computeranimation
Metropolitan area network
Front-End <Software>
Bitfehlerhäufigkeit
Datenreplikation
Total <Mathematik>
Gamecontroller
Reelle Zahl
Stapelverarbeitung
Kartesische Koordinaten
Gammafunktion
Bit
Mathematisierung
Versionsverwaltung
Kartesische Koordinaten
Zeiger <Informatik>
Netzadresse
Computeranimation
Metropolitan area network
Dienst <Informatik>
Rechter Winkel
Bildschirmfenster
Mereologie
Datenerfassung
Visualisierung
Gamecontroller
Bildschirmsymbol
Reelle Zahl
Default
Metropolitan area network
Filetransferprotokoll
Bildschirmmaske
Skeleton <Programmierung>
Vervollständigung <Mathematik>
Front-End <Software>
Fächer <Mathematik>
Wort <Informatik>
Gravitationsgesetz
Urbild <Mathematik>
Computeranimation
Managementinformationssystem
Proxy Server
Bit
Sichtenkonzept
Ortsoperator
Physikalischer Effekt
Fächer <Mathematik>
Regulärer Ausdruck
Diskrete Gruppe
Extrempunkt
Ausgleichsrechnung
Computeranimation
Vorhersagbarkeit
Portscanner
Metropolitan area network
Filetransferprotokoll
Prozess <Informatik>
Dimension 3
Softwareentwickler
Zeiger <Informatik>
Datentyp
sinc-Funktion
Klasse <Mathematik>
Gruppenkeim
Einfache Genauigkeit
Computeranimation
Eins
Portscanner
Metropolitan area network
Knotenmenge
Datenmanagement
Zählen
Wort <Informatik>
Softwareentwickler
Schwebung
Aggregatzustand
Offene Menge
Freeware
Open Source
Division
Zahlenbereich
Baumechanik
Computeranimation
Metropolitan area network
Bildschirmmaske
Verknüpfungsglied
Twitter <Softwareplattform>
Zustandsdichte
Zoom
Vorlesung/Konferenz
Baum <Mathematik>
Hilfesystem
Rechenzentrum
Umsetzung <Informatik>
Klasse <Mathematik>
Mereologie
Inverser Limes
Zellularer Automat
Wort <Informatik>
Komplex <Algebra>
Figurierte Zahl
Computeranimation
Lesen <Datenverarbeitung>
Multiplikation
Prozess <Informatik>
Rechter Winkel
Computersicherheit
Mustersprache
Implementierung
Kartesische Koordinaten
Quick-Sort
Computeranimation
Endlicher Graph
Metropolitan area network
Arbeit <Physik>
Datenfeld
Zusammenhängender Graph
Web-Seite
Computeranimation
Differenzenrechnung
Roboter
Red Hat
Computeranimation

Metadaten

Formale Metadaten

Titel Keynote: So, I have all these Docker containers, now what?
Serientitel EuroPython 2015
Teil 81
Anzahl der Teile 173
Autor Waite, Mandy
Lizenz CC-Namensnennung - keine kommerzielle Nutzung - 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 10.5446/20156
Herausgeber EuroPython
Erscheinungsjahr 2015
Sprache Englisch
Produktionsort Bilbao, Euskadi, Spain

Technische Metadaten

Dauer 58:17

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Mandy Waite - Keynote: So, I have all these Docker containers, now what? You've solved the issue of process-level reproducibility by packaging up your apps and execution environments into a number of Docker containers. But once you have a lot of containers running, you'll probably need to coordinate them across a cluster of machines while keeping them healthy and making sure they can find each other. Trying to do this imperatively can quickly turn into an unmanageable mess! Wouldn't it be helpful if you could declare to your cluster what you want it to do, and then have the cluster assign the resources to get it done and to recover from failures and scale on demand? Kubernetes is an open source, cross platform cluster management and container orchestration platform that simplifies the complex tasks of deploying and managing your applications in Docker containers. You declare a desired state, and Kubernetes does all the work needed to create and maintain it. In this talk, we’ll look at the basics of Kubernetes and at how to map common applications to these concepts. This will include a hands-on demonstration and visualization of the steps involved in getting an application up and running on Kubernetes.
Schlagwörter EuroPython Conference
EP 2015
EuroPython 2015

Zugehöriges Material

Ähnliche Filme

Loading...