Getting cross-platform: bringing virtualization management to the PPC world

so Ivory vividly sorry for the day of it's better now so am the it's my thoughts like this I will

just start OK so when my my

name is Omar Franco in I'm a software engineer and the median rent etc. the

threat and I'm also a over at engine maintenance and of attention project the government took about today is saying how we made

way to becoming a multiplatform management capable Starting with 86 and BBC 64 and basically how we got silver it to be in a multiplatform and I will fool them at the background of the the aim what we do to make this happen and a little bit about the people that did this war and a little bit about over for anyone who were less familiar with it and I discussed the problems that we had to face the solution that was accepted what we did I will show a little cold and configuration and then discuss a little in what is still left so summary on

our goal was bringing a multiplatform management capabilities to over in starting with x 86 and this is 64 and basically they have been overlooked by a multiplatform incapable why

why now so KVM empowering system was announced and listen to the and also in the was on the announcement of the Open pole consortium and by will IBM and Nvidia and some model the company so this becomes a relevant now in also it was a good opportunity to have infrastructure to have in a mall platforms so supporting over and just in case more platforms with support the view and the future so something very important in everything I'm going to talk about today was

contributed in by developers form and a lot of Brazil and the other is a not-for-profit organization they are located in those the land focused on that technology

ability development and basically these all community members there and contributed and the support and it was a really nice what we've that I want to talk about this just a little bit because it was very interesting process with these guys

so these guys who won the disintegration they sent their designed to the Oval tweaking and we make an hour another community did

some review to this design and some enhancements and once this design was accepted the development and implementation phase started and again we over IOC to make it the short-sighted because these guys in Brazil so was a big time difference and also over matters we sold out all the issues and our main panels will lead closely with these guys have been them in the figure out all the issues that they had in making sure what the code is being reviewed in in a timely manner and because we really wanted and we actually succeeded to make it part of their colony to flee fully solved so this is a really great success for us the OK so a little bit about all the

uh anyone here already knows all the to OK the so I go for it quickly there just to

make sure and everyone will know after that OK so by definition already is Laos gives centralized management and 4 so that that's the brutalization what it means is that we have an open source alternative for applications in like this sphere in the center and it allows managing the virtual data centers and the focus is on KVM if all this integration and performance will using KVM and there is a big focus on use of their use his of deployment for the user of it is really easy to deploy and use as an administrator law tool and create Europe and virtualized data centers and for users as well so the looks OK so that you can use over it for

a small little environments I like the cingulate under single holes running couple of yams and this is basically good for them on

testing because it here you don't have a some really important and every classification features like like migration no you can grow we've all heard 2 more depot 1 debate offender and we want to cluster

environment and basically you can see here I hope you can see here in over managers multiple that the center of each data center

has multiple clusters cluster and each cluster has multiple hosts so clusters some kind of migration domain and within cluster you can life might advance from 1 host to another and I want to give you a really quick high-level

architecture of full there just because I'm going to talk about this components later so I'm not gonna go too deep to and over so we have the engine this is a Java application runs of J because this is basically where all the logic and the relays and all the decisions are taken in our engine if it says scheduling decisions other decisions and this is also a In the gate all user requests are sent to the engine engine process it and and do something with it In the engine talks with the hosts and and that that that actually runs in the hands in the host with agent we it DSM the agent has a meaning a couple of important and responsibilities of fears he does on the host level configuration if it's storage all network and of course hand of within the political deviance on this specific cost will using a revealed for the VM operations if it's a start and stop in and migrating and we also have another package called mom that is responsible for scheduling and miscellaneous services that and finally of course on the host with a running a guest Sylviane so emotions and because they they're agent within a is there a package that responsible for ascending sending information from within the guess outside and if it's a IPE or applications that are installed of information that we want to show up in the in in the engine and what mean and also responsible for some months under any distance side effects for example see single sign-on that we really have to communicate with with inside the gates yeah OK so what was the idea so if remember before I showed you there are multiple data centers and each data center has cluster what you wanted to achieve is that we could have in cluster which is x

86 is today and also have fossils from water in a new platforms like class they gave PPC 64 cluster and in the future if available anything that's sports stadium so following is the question what and the goal of course was and induce support with minimal and as far as possible minimal changes it to day architecture inside the engine and to the you I to use the existing infrastructures and so on so what is the

problems that that we had to deal with the 1st overt was designed and developed with a single proffering mind and we only had 686 supported for so this is what it we developed for and this is the only thing that we

had in mind what what happens is that anything that was no platform specifications in for example the end devices in like networking displaying is just the same as in the physical work not all the configuration of supported on all platforms so for example and you know that you can

have this interface IDE via Tai overtire scuzzy and in we found that the persist 64 doesn't support IT so we had to do in some filtering and to block users from trying to use it on BBC because we know we call for the them more problems that

we have so many assumptions 1 made and without taking a lot from in mind so for example the site

addressing is different between x 86 and BBC 64 so we had to deal with the end in runtime we had to change the code that does that address and I will show that later also some features I am not supported in all platforms if for example like migration is still not supported in PPC 64 so again it's a feature we have to block according to the architecture so the solution that was suggested then eventually

and we used was using this strategy and design part and I have some In diagram of it basically and the strategy designed part what

you say that you have the interface of any what is supported what are the actions that are supported and you can have very different implementations so implementation 1 and 2 the implementation for x 86 and implementation for pizza 64 what is allows us is

in a couple of an important things 1st selecting the behavior at runtime and we don't need all kind of spatial beliefs and the cases in the cold to select which go to run and eat Gödel's there in specific code to want to get it could become encapsulate 86 gold and index 86 started and the PPC coding the pizzas but the gene and the

fellows and a people that are entering the cold tools find in the specific called very easily because people encapsulated together and a very important is that it allows us to use the word in any mall architectures in the future In other

stuff that we had to take in mind is we are in defining a architecture in the cluster and so we could have In 1 data center different clusters any platform so you could have as they should before class TOEFL PPC unclassified

66 and distribute type is appalled by the host it's allows us very easily to decide if a host can run in in in the cluster or not and also in over 2 we have 2 configuration packages that are in pain we're using it as infrastructure for deciding in east of some features supported so 1 in the 1 thing we have to do is add the architecture awareness to the configuration so 1 is the global configuration s'll features and on other and a package that we have we call it always sinful it allows us to any manage and capabilities in the guest OS that so in for example Windows has a set of the devices that the data can be used and the so that different set so we're using we are and saving it all in 1 configuration and what we have to do is to add architecture to this configuration as well and on you will show an example in later OK so what was done so far we can divide the tools slim main phases 1st was to identify more specific called in to the distorted you classes in here than any changing of the configuration and in creating the new configuration for new architectures that we just added and then finally there some specific coding folded because 64 so m we had to

as I said at the architecture field and to day in a cluster and also supported CPU as we have the support it's you

list in over that we are in supporting and once we implemented the strategy design pattern so we could move all the specific called a to the the right place and basically create some consolation for the lexica success specific configuration

so I will show I think it's the next slide I will show you how those in configuration looks like and it gives us in great efforts ability because in the contacted

the called didn't have to invent pretty much anything they were using an existing infrastructure lsle they basically just add the new architecture and the new values for the guest OS is and configuration and it was just walking out of the box gives us the flexibility in the y forces during the we show that as well so this sourcing for how does it look like

so a this is a snippet of their 1 last and you can see that

we the PPC 64 bit OS and we added the CPU architecture in class we being and the value for this OS is because 64 so basically that is either OK so on we have to do is a and x 86 to them in to the base OS and of the existing last and just this 1 and on new always says that we are going to add a 6 and so on you will inherit from this OS so basically it will be really easy to add more supported OSS once we verify our what the and am also you consider 2 here we defined the devices there are less so therefore the specific OS and since it's a it's a 64 and then display protocols possible we are only in avian saying and because spices not supported on BBC 64 and this is how we use it in the cold so although sinful just you and ask for the package give me what display types are supported for a for this OS induce come in this compatibility version for the cluster and it will understand from the OS what these and they architecture and the times there in a list and we show in the I how to extract as well so specific code that had to be done so 1st is the specific addressing that they took before and because the

addressing the PCI addressing In the PVC 64 is different than x 86 and I will show that also aware that this specific at PPC 64 a devices for the on and discussing this opera or how you pronounce it and a lot of front-end adjustments and I will show you right after that they to filter out on the directive on 2 in choices for BBC and we have as I said before to block unsupported features if it's uh snapshotting the and migration In the DSM stage and that I mentioned before there was some changes in the topology and whether the 1st assignment for the reported information and also how the information that had to be change according to the new people so how does it look like in the

cold and you can see that and 1 really this is the code before when we were building VM drives so there is no ask about the platform here this is just an if it's the scuzzy this is what you gonna go there is no in addressing code here because on x 86 and we were relying on a automatic addressing from we were relying on automatic from the very and indeed this is 64 we have to be aware of the dressing so is it clear I hope it's not too small them you could see in other right under the case there is need and the if that's a OK there is no water utterance so let's credit and basically can see here we are getting the strategy by the Vienna Inn yeah Waddington The strategy by the VM architecture so this way and the code is not aware of what what architectually sign in like now because will get in the lights but to gene and then money in the cold so basically 1 and there are signs because the other school so basically they're the cold doesn't is not 0 well in all of the specific architecture because of the strategy they were also headed visitor design quadrant in which has the specific and implementation in the sub project

and it comes in handy anything the strategy gives you the correct and code for day and architecture and the visit or any is responsible for actually executing decoded for the specific command if it's 4 are the same and so on and now we show an example so you can see here that we have are

signs scuzzy and it's architectural command and we have implementation full full x 86 and implementation full PPC and you can see that for x 86 it's still empty because will still the line of the old ID Louisville and logic and for PPC and we have different addressing the the civil post cause the author or if it's less popular discussed so the controller diversity and the component is different so what is we have ready and you can create

clustering mentioned the architecture of yams templates and pulls everything that over has and importing and exporting keeps the in architecture and information attach induced so according to their specific and strategy architecture storing and searching and we can search VM's by the specific architecture and managing the answers his starting stopping of course so a little bit how it looks like I hope you

can see OK so this is the new cluster of dialog and you can see that and below we had the new supported and

CPU so it's all in family of IBM Power and basically once the user selects and the specific thing you then the of their architectural is In set from disappear so every every CPU as architecture so it's about this is then you virtual machine dialog so you could say

that of all the clusters of PPC cluster and so the operation system is filtered out and I can see on the the uh supported so can't you only have order O S about and once we have verified more avian losses thank you this list with full show below and this still then you virtual machine dialog you can see as I mentioned before spices is not supported on the PC so again and in this is in the U I and this is a really nice because it's pretty much magic it's all being 1st thought infrastructure always sinful does this for us so it was really an easier lasted

and a new disk and therefore VM of so again you can see that uh the this interfaces all of the adult local called into the architecture so we know that we are adding these 2 to the end which is the PP CVM so we will freeze felt idea because it doesn't fall

what still need to be done so there is still some issues with from network on PPC and send this the guys from rather or chicken the solved and what more important we have looked at features on PPC because there are still not implemented or not exactly walking so like migration is that in most important 1 also and taking snapshots the N snapshots still not supported and hot club for this is not supported we do have a log plot for makes full network interfaces but the this is still not OK so some of and

today thanks to this work and over engineers with apart from ready and today we have x 86 and PPC 64 would being in that flea for release and but it's very important to say that we now have the infrastructure to is the add in more architecturally so once we have supported on more architecture it will be should be is a walk to the and there's a lot more information sold to resolve website and the wiki that the guys from Eldoret across the the level created in for the engine for the media cemetery in projects and all the information on implementation is there that's it yes

the use of the and so on of and so the yes the fact that so basically it would have difference between this CPU because there ever transition features of different KVM is different on this uh all on this versions because you have different features that you could use I don't know the IBM features by hard but let's say that once a few as as some features and the Caribbean can use and the other doesn't so I have to specifically say what is this a you for this cluster because the want migration toward our so basically you don't have to fit this or if you leave it blank and the 1st post that you will add to the cluster will set this for you yeah we have automatic setting the if you set it if you live with them yeah and the best yet what it is that the 2 of you I I think that there's some issue there with the storage I don't know exactly what was the problem with a snapshot by so so the on BBC so maybe I'm not updated I I hope it's useful in and the we have to do to do integrate all in over did you integrated in the work as well so they are also this they also ignition warming so the war concave M and was announced in regardless and I think they did some work in the rigid but mostly it was really a 7 the engine so I think that very thus supported and what they're all issues in invidious but so therefore it works better on over than in the cells sorry it works better than over time in the and when I just use blame the word with the social order with the regions of the fitness for men they ask you if that's the case at the on the it should be I should war will worry is inhibited the same as is in very so we this is is an inverted as well OK so it's in literature there but I have a question about the news that are on yeah you know I'm quite morning on the of to and I can't wait till March because on and that's at about a new form of 2 rules because you know of any work in that I have no idea I have no idea sort of you know if they if they were going kvinnor thing was all I don't think so but I didn't look village OK yes right time sorry again in on and thank you and what 1 OK so the question was actually at this point what you're displaying or what I understood from your displays that you already have Linux running on a machine and then you can build other Linux is on top of that yeah that would mean that the Alpine had already be defined is there any plan was difference for the ELP on the logical partition on the machine itself OK so is there any plan to interact with agencies so you can do some configuration of the of power do dynamic Alpa's on it I actually I don't know I don't really in a reasonable assumption on IBM porous so the I think it's really because we were using the vehemently is his user space so I don't I don't think it's too has any OK we have and then you're much few