 All right, I think we're on Okay, so we're gonna talk today about both a new model of cloud computing that we're sort of pushing forward as well as a new Instance of it the Massachusetts open cloud So just I know this is obvious everybody that's here But you know clouds a lot of people are asserting that you know, they're gonna have incredible economics You can actually spin up VMs or you can spin them down And so for the user of it, you know You only pay for what you want and the provider meanwhile has massive capacity that he can put in there Where data centers are cheap, etc. And a lot of people believe that because of that public clouds are gonna kind of dominate the industry that you're Not gonna be able to buy a computer in a decade from now We don't believe that's true We actually think that the current model of cloud computing the public cloud where you actually have a closed cloud You can't see inside it can't actually meet that vision a model where a single provider is actually standing up The cloud is gating whatever people put into the cloud where innovation has to flow through one company or one organization Is fundamentally not going to allow for the public cloud to actually meet that goal And today you see that technology companies are largely locked out for them to involve themselves in the cloud I mean selling to one big provider Which is a huge challenge and also Performance sense of applications are really hard to get right on a cloud when you have no visibility into the infrastructure And then we have sort of all the other issues that people know about of you know Even vendor lock-in by price is it really more expensive to move a bite out of Amazon than it is to move a bite into Amazon So we think a new model is possible a model where not one company stands up the cloud But a whole bunch of different companies can actually stand up different parts of the cloud Multiple people can stand up compute multiple people can stand up networking and storage and they can actually be in this environment Which acts as integration lab and access A place where you can go to a diversity of offering both competing and offering fundamentally differentiated value And when you look at this world at first you sort of see well geez you know Is it could be complicated today? I go to you know public cloud and I just buy VM And it's a really simple story. I might have two or three flavors in here where there's different SLAs where there's different pricing Where there's different characteristics and accelerators Well, you know what what's the user going to know, but the truth is Just like any complicated marketplace what you end up getting is intermediaries you get up Applications whether big data platforms or SAS platforms or past platforms that actually can do the investment to look at the underlying Infrastructure and find out for their customers or for the applications. They're running what the right kind of environment is So instead of having that all in the plumbing which by the way becomes really complicated the infrastructure level to build a complicated Enough constraint engine that understands affinity anti-affinity Accelerators locality to data Regulatory constraints like HIPAA instead. Let's expose the pools of capacity and let people on top of that build intermediaries that go against that capacity So that's kind of the vision of what we want to build I'm going to run through a couple of quick examples And I'm going to turn it over to yon to kind of walk through the pragmatic side So first just kind of looking at what we do with cloud today Instead of having one company standing up the cloud Imagine that we had a cloud with a lot of different companies participating in it and this is kind of a cool use case company called community space which is sort of a mid-sized company in Massachusetts and you know They actually are in 150 countries. They have data streaming in from all over the world And they say well, we don't really know what to do with it You know and we could go and build an in-house data center and have vendors try to tell us what to do But actually picking when in that kind of arm is really difficult So wouldn't it be cool if you know, we actually had the mass open cloud We can actually put our data sets into it or at least representative data sets into it And maybe we run big hackathons on top of that data and maybe we first figure out. Okay, cool Here's an application that can use space wants to use all the time that goes and you know Is actually giving them back real-time data on how their environment is working And maybe that runs on spark which is a really cool big data platform and it runs on top of you know relatively small Environment like you know cloud in the box well now they have visibility to that They see what it's running on and where the performance characteristics are right and they can actually choose to in-house that which is great They can use this just as an environment to test things out But maybe there's another application which they want to run once a month It runs on Vertica which is a column oriented database that requires low latency communication So maybe the Plexi networking and a massive SGI rack of thousands of nodes and they only run this once a month And they get deep strategic information about it. They're not going to in house that But maybe you know community space isn't that happy about going and using a cloud over the public internet You know, they're just not used to that environment Well, it turns out that Juniper has a really cool solution for extending your private network from say Verizon for their enterprise Customers into the cloud so as far as you know community space is concerned. It's just an extension of the private network Now this is just representative But what this is saying this is saying like 10 companies are offering their value to one mid-sized company They couldn't sell to it. It's the innovation of a lot of different companies providing their value through a shared cloud environment Just briefly kind of walking through a few other examples, you know cyber physical systems interacting with the real world There's a button team that's kind of thinking about how can we actually do that where we could actually get a police car going from You know Newton to Brookline to Boston Full speed through and turning all the lights green before them with actuators and lights and sensors all over the city Well, we can do some of this stuff But first of all the data right now is actually divided into like 10 different municipalities So we need a common place where it all streams So cloud kind of makes sense but on the other side the vendors that are controlling the actuators really want to have their Own gear in the cloud. They're not going to trust to putting this inside some third-party cloud where they can't actually put their Own gear into it Hype forms computing we already see a convergence between high-performance computing and the cloud We see that convergence happening because of things like, you know full bisectional bandwidth networking that's just all the big public clouds are going to and you see There are big HPC clusters being stood up on top of these public clouds But if we actually an open cloud we could do a whole lot more you could imagine actually mixing between having HPC workloads and Non-HPC workloads and moving the boundaries between them So you know today most of the elastic clouds run at about 50% utilization if we had the capabilities We could drive up the utilization to like 80 90 percent and get enormously better economics Also, one of the things we're working on is get massive at elasticity So you could imagine getting thousands of nodes for a minute and what applications you could get through that So for example, there's this neat runtime that a team at BU's building where you have a Linux front-end node And it kind of spins out a whole pile of nodes and then collapses them away to run an application These are all written library OS's that can boot bare metal and we're actually targeting this as a medical Imaging application so you have today. They're doing fetal medical imaging at children's and you know You can't tell the fetus not to move so you end up these really fuzzy images and people are making life and death decisions on these Well, you can actually take this and segment the brain have a human being go and circle the right areas and then actually do Do 3d reconstruction and motion correction on it and takes about 24 hours on a high-end machine So it's clinically irrelevant for people making life and death decisions There's actually a team that took that took the core part of it and spun this out over 10,000 nodes and we're able to solve the problem in one second So this is representative of class of applications, you know, if we could get that elasticity into our clouds You know and you're not going to get that by an existing vendor saying, you know, this is a use case I totally don't know that that doesn't even exist today But if you can actually have a community of people working on a cloud, that's the kind of thing we think we can do Last one is you know just an example of a whole class of research things So today clouds are priced based on stickiness trying to get you into the cloud and lock you into it If you actually reflect the cost of the cloud into how things are priced That's good drive efficiency because now there's a relationship between what people do and what's it's actually costing them And you could do things like represent the SLA and the characteristics of the application and map that to it to the Infrastructure so, you know, you can actually drive the real price of it to be represented of the workload And so if for example, they could be squished they have an incentive to be squished because they're paying less So there's some really neat research that's being done there Nobody can deploy that an existing cloud because it's kind of locked behind a door stood up by one provider You can experiment with all these models in an open cloud and At a fundamental level, there's a ton of these research projects going on All of which are highly academic have no relevance the real world because they can't do things the real cloud And I think a lot of people would feel geez that's not so different from a developer that wants to do an at scale offering But actually can't get access to a real cloud to do it or do anything at scale So it turns out we we when we actually proposed the talk we weren't sure we could talk about it I'm gonna hand over to you in a second, but we kicked off this project the Massachusetts open cloud This shows it's it's a joint cloud. It's being stood up by the five big universities in Massachusetts So BU UMass Northeastern MIT and Harvard In top of the big shared facility we've built and it has a broad set of industries So we've got about 16 million to 20 million and in matching commitments from industry and The week before last develop Hattnick stood up in front of the audience and actually announced that the Commonwealth was supporting this cloud and Sorry, it was 30 seconds over You did very well I need to see if I can operate this high-tech device. So yes, I can excellent So yeah, so as Oran said it is finally official We were kind of scratching the curve on when we can announce it officially which is great now The state of Massachusetts behind the project It's being actually located in a brand new data center in Hollywood Which I will talk in a minute about which is a green data center just being built from scratch in 2012 built up in about 18 months Massive scale massive opportunity to host a lot of new equipment using some really interesting Technologies around cooling for example with chilled water loops Hydroelectric energy for example so by commercial standard might not be a super mega data center But it's still very large with 15 megawatts for example two acres of compute space And it gives us a great environment to host a very Variety of workloads Installations computers equipment, etc. There's all the whole operations behind it So Harvard University will actually operate the entire environment So we all have people already all very very worst on for example operating large scale HPC and cloud Environments, and that's just sort of a picture on well This was the beginning in 2012 and then magically into 2013 February 2013 The data center was ready people moved in Harvard BU others and within two months They went from ready to production workloads, which is pretty amazing for a data center to see and this is just a look inside the data Center where you see basically all the wiring coming from top the chilled air loops Water loops for example on top here. You see like for example the aisles of equipment in there today So this is a really really cool data center for us I thought this was actually a great statistic from last year literally just after they went in production in June I mean I got this chart from James cuff from Harvard where basically they gathered statistic on how much power they for example Consumed just for the months of June and you see here Harvard for example consumed 199 million watts Just for the months of June. I would have loved to show another chart, which was kind of nice I talked to Oran yesterday where in July New England ISO electricity provider actually sent out the memo to all of the large power Consumers to conserve power because of the heat for air conditioning and James cuff actually tweeted a picture where he showed we shut down All our jobs in the data center and literally consumption went from yeah a hundred million watts to zero in no time So he actually had like this chart where you see like all the way up and suddenly like just flat line Which is great and shows like the flexibility They have in terms of scheduling jobs and all the managing the data center, but enormous gross opportunity and running at scale today And clearly open stack is a great technology now to utilize for what Oran? Explained Massachusetts cloud to build this sort of what we now coin this Multi-land-lawed cloud environment where you no longer just have one installation of open stack for example And then you only have one choice of service But have multiple services we compose your own service from and then you get to pick and choose based on individual properties An open stack gives us a great base for this because of its openness modular pluggable infrastructure already existing API's scales really well for example and a very very vibrant community However, there are a couple of areas and I will go into them briefly. There was a talk yesterday By our partners in crime Peter Jay and Brent. I'm going into much more technical detail around the issues What needs to get done an open that can actually accommodate such a multi-land-lawed cloud, but there are a couple of areas where Today opens that hasn't been designed to cater for environments where you want to pick different flavors for example of services So in an open cloud exchange for example, you may have different Instanciation of different serve of the same service you may have multiple novus multiple glands for example Multiple keystones different network providers different as the end solutions and you would like to pick SDN from over here Nova from over here for example and Combine them across different open stack installations and today that's not possible because of the way for example open stack Not only identify services, but also communicate secured services For example also how some of the API's are being documented or not documented So we go through some of those issues, but fundamentally they're not like major showstoppers We feel very confident Open stack can cater for them and actually one really great experience here The summit has been starting some of the early conversations with the community and upstream everybody actually was very excited about the fundamental model of the cloud exchange and Everybody actually agreed it would make perfect sense and actually or and had a couple of really good conversation made great Headway already around a number of topics where we thought it would take much longer to convince the community actually do Summing and they said oh no makes perfect sense. Why not do it? So one example here is for example, how do I pick a service where as Oran said in his section Where now you have different offerings maybe from the same service type a cheaper one Bob's bucket bite For example and somebody needs GPUs you have for example a more mature provider offering services But how do you find the right one? For example, how do I connect to it? For example and today? That's a problem But part of this is the goal is all to build a marketplace on top of this where you as the end user or consumer Would for example able to just pick Services being it storage being a network being it compute and then just compose your service build your VM And it just uses for example The service underneath but you cannot now all the imagine not you just picking yourself the services But actually having intelligence in there. We just specify. I don't know an SLA a cost factor You want to achieve or not for example? Over go over a cost factor and then barely just build these services automatically So this gives you enormous choice really interesting opportunities here And gives you again the ability to cross multiple environments. No, there we go So but with open-stake one area is for example How do you assign or how do you? Automate the environment secure as you give for example access to an interval user to a service today The assumption is once a user has access to a service They can access everything within the environment and you want to make sure you can actually separate these And therefore be able to also reach Services Individually today for example Nova will give you you ID and the you ID is from end to end You can't identify individual services individual Components and that's something which for example would have to be broken up Another part is securing internal Communication because the assumption today in open-stack is anything within an open-stack installation usually is on a private Internal network and the components come communicate internal like Cinder and Nova will communicate Together within the same security framework today, but in the new model with the cloud exchange You may have Cinder in a total different environment than you have for example your Nova instance You want to make sure there's secure communication and encrypted communication. So We are looking at ways. How can you first of all document those internal APIs or need they need to be documented? So people can actually separate these services and how can you secure for example the communication between these services? And that's not only Nova and Cinder this could all to be for example console access This could be also on the SDN side on other components. You can play this anywhere In the open-stack framework here Then also, how can I use untrusted services in the context where today you might request for example an instance? On a Nova node and you get your Cinder volume Which means you have access to in theory to all of the volumes part of this Nova compute node, but in this model here with the cloud exchange You want to be able that whoever has access say to the blue disc Own or to the blue VM only has access to the blue disc and can't for example ever get to the red disc For example and vice-versa you want to be able to have service segregation and again remember these discs could be in a different Open-stack environment So you want to make sure that again it also spans cloud spans infrastructure and again You don't end up in being able to subvert data subvert VMs from other environments for example Encryption I mentioned earlier for the internal services, but this is now also applicable for other Services today already used for storage for example So if you look at storage use ice-gazzy today the communication is not for example encrypted between say the storage back-end and for example The Nova side so we are looking at ways to what can you do to for example secure the storage traffic Between those instances being it via VLANs being it via other technologies So that's another area to explore and I think there will be a great opportunity to collaborate again on all of these with industry partners to figure out ways what are safe ways for example to Secure communication not only for the storage side, but also on the networking side for example. How do you then communicate? between clouds Today again if you use for example an SDN solution the assumption is it's in the context of a single open-stack installation But how do you bridge for example multiple open-stack? Installations and then get the ability to seamlessly for example communicate between them and maybe even move workloads between them as well That's another area. We need to look at as well And today for example if I just quickly go back with SDN is today you can do this Maybe with one vendor, but also how do you get multiple vendors for example to interoperate? So that's I think a really key feature of this cloud exchanges to get multiple vendors engaged and then it Interoperate between them not just within one vendor stack, but across multiple vendor stacks So here would be kind of list on what we see Where some of the benefits and all that leads into why for example, we actually participate here But one key feature clearly is open and multi vendor environment So today and I can say it is coming from redhead is when we do certification or testing usually We build one environment with the vendor together do the certification break it down done But today nobody really offers for example an environment where multiple vendors come together can do cross Certification cross interoperability and make sure their gear actually works together across boundaries as well You can look at for example now building new and innovative services around this So you could come in as orange that earlier vendors could come in and say here. I've got this really cool new technology I would like to put this in this cloud environment and then offer a service and customers might gravitate toward based on the properties being At performance being at capabilities being at price for example and then also the ability to Spend across multiple diverse environments So when we talk about cloud usually everybody thinks about the virtualization immediately everything is a VM But we actually see more and more Interest in for example blending high performance computing into say an open stack cloud type environment And now suddenly have the ability to for example mix and match workloads within the same environment Deborah for example driving utilization up, but also for example creating again new interesting opportunities here All the ability to look at for example per service pricing So today when you go into a cloud you really get the price for a cloud But here you might be able to pick again based on price for individual services rather than for the entire cloud And then also looking at new ways How do you actually orchestrate and compose? Workloads and Applications within a cloud environment. I mean we all talk about how you can in Siri build these multi-tier applications Where a database tier runs on one side middleware somewhere or middle tier somewhere else endpoint somewhere else today That's not really easily possible Besides on slides, but in reality this would give you an environment actually can build these Environments and also get the community involved for example have developers explore the real use cases in a real-life environment here And then couple other benefits here is it's not just the technology showcase but also again the ability to collaborate within a diverse environment because Today a lot of the collaboration happens behind the firewall often in very small environments Our developers often crave to have access to real-life environments with real-life data real-life workloads Especially developers here today would crave to have access to real-life open-stack environment with real workloads Real fair scenarios, but without the pressure of being in a high pressure production type environment So here you would get exposed to operational data You get for example, maybe to build a new service in parallel But you can again study this in real life look at log files look at ways for example to correlate events in such a highly Diverse and complex environment Look at new troubleshooting technologies because again open-stack now becomes very complex as we already all know But now imagine this would be an environment with thousands of nodes suddenly things like event correlation How do you actually mine log file data becomes really really interesting and really important? I think there's a lot of interesting opportunities here and again It's not owned by a single vendor, but really everybody is part of this effort and can again then Participate bring value to the table beyond all the just technology, but all the people So I see actually a real big value here, and that's all the truth For us is it's not just a technology side But here you can build up skills foster new skills for example in a real-life environment where otherwise It's really hard to get people engaged also get people for example to participate in the communities and build new communities around For example some of the projects in this cloud How many people here are developers How many of you keep your hand up if you actually have an environment with a thousand hosts that you can actually test Okay, so we're building a cloud and one of the developers in the room actually can do a test their experiments at scale Sorry. Yeah. No, no, that's fine And I bet the thousand nodes installed from the gentleman who can test might be eventually half at least half production So it's not like you can just jump in and do stuff So that's actually a nice way where you might just be able to carve out for example Areas of interest and then do some exploration work and fast and we've started to do this We've been working with MOC for quite a while and fast the real interesting aspect is again to Bring people into the environment work with people like Jay Peter Ian and Oran to actually get Operational experience running a real cloud open stack environment at scale and be able then to build their skills Which they can later on for example bring to customers so they can drop they don't get dropped into customer sites like and for them That's all new they actually have a lot of experience just operating this deploying this and now he can play this further There might be opportunity to come up with some creative internship model or whatever down the road as this scales to kind of give people the ability to That kind of do on ramping for open stack deployments or even if you just want to get used to open stack here It also allows us to for example Have support services from our side treat this environment like a real customer So the MOC folks can call our support people like the real customer, but our support Personal on the back-end side gets rotated through this role for example So we end up in training support people all in understanding large-scale cloud environments And then for us, it's all the interesting to look at again new workloads new use cases large scale experiments as Oran said earlier All to the ability to maybe leverage some of the other Products we have around Jbos like rules engines Jbos data grid you name it There's a lot of opportunity not just fast for every vendor to actually how is this applicable to for example then manage a Multi-cloud environment multi-landlord environment And also the ability to do some very early testing about around new technology and Hardening of new technology at scale because all here have been at the summit all week And if I would ask what was the number one issue number one topic I would bet there was probably something with an end at the beginning Network and neutron and this would be environment for example, you could test stuff very early at immense scale for example Same would be true for deployment tools So J actually Jason actually looked at some of the deployment needs for MOC around bare metal deployments in the context of high Performance computing and the good thing is here at the summit It turns out there actually is a lot of interest on the developer side to actually solve these problems So Jason actually went out and built this project called harsh hardware as a service But it seems there might be actually an opportunity to make this interoperate with other technology such as ironic and again It is a perfect example to like how do we can how can we collaborate and how can we work better together? And really for summary I probably give it back to all right Like what do we what he also wants to get from you in terms of collaboration and feedback? So I mean the takeaways I want people to come to walkways We're not we're obviously this is a place where we can refine the software that we're deploying in private clouds And that's gonna be a really important problem for the foreseeable future but we actually think the public cloud can be really important and When we move away from a model of a single company standing it up top to bottom Controlling and gating the innovation, but we provide environment where multiple companies can get involved and in doing so that's an environment where Developers can get involved where researchers can get involved So a kind of open cloud and we defined a mall we call an open cloud exchange And that mall of an open cloud exchange has a variety of providers And it also creates a marketplace for people to create intermediaries to go and consume that underlying capacity on top of it Open stack, you know not developed with this in mind Developed as a multi-tenant cloud, but not a multi-landlord cloud. It's got a lot of it right. It's actually pretty close But there's actually significant changes that that we're gonna need to do to enable this environment But the neat part is and we've seen this verify time-to-time again the meetings this week Is that actually the changes are all good for open stack? Hardening the internal interfaces having them go through the same rigor as external interfaces Identifying ed points it turns out that in our conversation around cells They actually satisfy a bunch of what we're already doing for reasons that rack space had to do to offer a diversity of services Now putting them under multiple providers and taking that same model and saying no, that's not just for Nova We actually want to do that same thing for Cinder and then putting that first-class, you know delegation You know fine-grained permissions none of these is rocket science, but we have to actually Get this in we don't need all these changes next week to enable this model will enable its service by service In this environment, and we don't need all the trust issues dealt with at first. Okay for the first couple of years There can be some degree of trust about people But we want to have a cadence of putting in the features and changes we need and most of what we want the community to be aware of them You know when they're thinking about deploying services and open stacks or doing new development open stack We want them to be really cognizant of this model in which we want to start deploying open stack We also as Jan was saying really well This will actually offer broad value we think to the community a place where people can come in and do at scale Experiments where we can have for example brocade or Cisco set up SDN solutions alongside each other and You know and actually be able to evolve their offerings very very quickly But be doing interoperability test between them and figure out how we can actually go to this SDN solution And this SDN solution and have a single tenant be able to create VMs on them and create horizontal interfaces Say with open daylight for communication between them So we'd like to invite people to get involved That's it probably just for Questions if you probably can go to the microphones would be useful. So everybody can hear the question Yeah, the race is on I think you won So I'm sorry if I missed it, but I mean how are you gonna be managing federated identity or That kind of stuff are you actually going to have like one identity provider for this whole thing? Or are you actually going to do federated identity? And if you're going to be it sounds like you're going to be doing like the broker model on top of possibly different providers but Is there going to be any notion of doing federated resource management both at the infrastructure level and then also At the application level because you're gonna have all these users and teams or whatnot that will want to collaborate But they may want to do so You know using application level tools as opposed to just the infrastructure services I think I count four questions there But so let me kind of touch on them and some of this is TBD But the first the first point about federated identity Absolutely critical requirement now we can we don't have to do that to start off with right because we can say hey We're all trust me, you know the MOC is the one trust the entity here, but that's the wrong answer So we've been looking at what's going on federated any we're really excited and we want to exploit that We didn't identify there's a problem because there's actually an active work going on there But we we need to actually be one of the use cases that people consider as doing that development When we start talking about The broker I think that we could do is so if you look at the way cells work, right? You actually have you know a nova scheduler on top going to different cells and doing that brokering and that's absolutely fine When that thing understands all of the possible applications that everybody went might want to ever do to the cloud and all the possible constraints at least in my previous life building a cloud for another Company is that it was really hard at the infrastructure to get all that right So our model is that yes, we can build those things and they're relatively simplistic models And we will and we can actually use exactly what's in Nova today to do it for example for Nova Scheduling we can do that, but I think that if we want to showcase diversity of things our mall is instead Let's actually demonstrate show the pools of capacity and then build this in libraries that could be incorporated to applications So those applications that understand their own requirements can go and pick Three of these and four of these can say okay. This is my critical database I want it to be on that highly available storage and these are my web servers I'll use this kind of storage or maybe this has to be HIPAA compliant So they can actually make that decisions We don't think it makes sense that there should be a schedule that understands the world We actually want the scheduling to be moved up Yeah But so but there will be a resource discovery Issue around that such that you know any user application can find the right things that they need to make right So we're actually one of the things that we didn't describe the solutions at all We described was the problems, but there's going to be a rich services directory that we're building in here And that service directory will tell you the pools capacity and our assumption is each of those pools are relatively large So you have to know these services exist. Yeah, but you don't have to you know allocate every individual node You just say I want to go that kind of over for that kind of resource or that one for this resource Yeah, I think you have a great sandbox. I think it's gonna be you're gonna have a lot of fun I work with HPC resources for academia for researchers you touch very briefly on a Finding novel or innovative ways to integrate HPC in an open-sac cloud environments I'd love to hear more about that and maybe some examples Sure, so I mean a one extreme it's you know if you look at say the North Carolina VCL project What they did was really cool. They ended up sort of saying okay This is a private cloud solution It doesn't have all the bells and whistles of OPE stack that's been there for a long time satisfying their enterprise And they actually bought HPC like gear and then they don't use open-sac For actually the HPC applications. What they do is they use a standard scheduler MPI whatever and they move the boundary between the different kinds of resources and That's great because what it lets you do is is drive up the utilization because there's an infinite number of HPC tasks all the time So it actually gets much more efficiency But you know the super elastic applications like building into glance the capability of bursting of Broadcasting an image to a thousand nodes so I can actually deploy like that library OS. I was talking about We're already moving we see an HPC we see HPC for a long time. It's had full bi-sectional bandwidth networking Well, we're actually doing that now in clouds, you know, we did class now a class We organize our networks in the class thing we use multiple stages of it Using combined networking, but it's not obvious that makes the most sense if we're moving to a fully integrated You know close to data center scale machine or large-scale machine So I think that that HPC till now has been a niche in research universities big national labs HPC can be part of our daily lives if you know You have applications can burst out and get a lot of capacity very quickly and we think that's one of the potential things We can do on a cloud so you can support legacy by moving the boundary But now the next thing you start thinking about is HPC applications already have to be tolerant of failures once they start moving up to massive scale and So bringing new programming models that actually tie into the cloud is an obvious kind of thing that's going to happen. I I don't know if you're aware of Larry Peterson's open cloud initiative. He's got like 20 internet pops with Folsom running and He's looking to recruit in a bunch of planet lab nodes and racks at various different universities currently running Folsom and He it's I think it's open cloud that net something like that. Have you checked this out? Yep. Okay Hi, I have a question I got in late so the you probably already covered it and because I read the description and This is an amazing model. Obviously has a lot of great impact and you know, especially the social and to community but on the the more commercial and economic perspective and I read something in the description like the participants Especially the companies and I guess commercial entities can also derive their revenue somehow. So they're like a common You know kind of a federated or you unified billing charging system And so that people can leverage on top of that, you know, based on you know The resource they contributed and then I get the you know revenue. I don't know if that's already covered Well, we didn't talk about it. And you know, we it's absolutely critical to this is where you know for everything It's deploying open-stack will be using, you know, standard metering techniques and everybody should be metering their own offering You know and the idea is actually you can go to the cloud You can buy the service from who you want to And the cloud the MOC is the thing that's going to be responsible for actually the billing But you know the metering of it and the determining the rates is actually what each individual company should be doing now There's a lot that has to be involved in the business model the first year We're kind of offering this for free to the research universities with a limited access to the public But but that's actually our three-year plan. We've promised to come with within the third year I think by the end of the third year, but sorry the end of the second year We'll have a model that that will be implementing in the third year It could be something like on top of salameter or maybe some you know add on other, you know Vendor capabilities and well our basic assumptions is this is all built on top of salameter But we've got enough problems to worry about that. That's another year or so that I have to really start worrying about that. I Think one more minute one more question. Hi, first of all, thank you very much for the interesting Presentation and for the interesting topic and the question So what is the definition of the cloud in your model because from a first look it looks like that's your Building clouds on demand So it looks like you have like cloud of clouds So in this model, what is the definition of cloud a cloud of clouds would be a malware One person stood up open stack another person stood up open stack And then we did some sort of federation between them and that's that's not the way we envision it This is actually is probably the way it's gonna start But I want a company that Understands how to actually do storage and a really cool innovative thing maybe solid fire something to come in and just do Cinder in here Right and actually manage it and be able to iterate on I want pro-k to come in and actually have a neutron maybe with maybe with Open daylight under it that they're evolving really rapidly and deploying in an agile fashion on top of their switches and stuff I want you know EMC to be doing its storage and I want people to be able to mix and match Maybe HP is doing everything in one one one one step But I think people want to have their data come into the cloud and then they want to be able to move from one kind of compute to another Depending on the characteristics or they want from the same compute to be able to have storage from different things So I think the way we think about this is take what open stack is today a set of services well-defined API Each one is really a separate service, you know That's kind of Amazon's mantra is that all interactions go through the external API, right? So just harden open stack and so each of these things can be stood up by separate provider that really is focusing on their area and Can innovate at their own rate and isn't dependent on all the other parts of the cloud is can offer their service Does that make sense? Okay, I think we're supposed to wrap up now. Thanks guys