 Good afternoon everybody. Thank you for being here last Session of the day, so that's great. I love Boston I'm very happy to be here today And I have two reasons for that The first reason is that Boston was the first city in the United States of America it visited 20 years ago I came here to study English for a month and hopefully all of you can notice my Boston accent Second reason and a little bit more important is that today we're introducing Hyper-SQL for open a stack. It's a so we define a story solution that has been built for open a stack cloud Today, we're going to cover how hyper-skeletal for open a stat can provide quality of service performance and data protection build on software that can be used with the commodity hardware of your choice to help customers to move more workloads into open a stack clouds My name is Carlos Carrero and a product manager working for very transfer the last 16 years Today have the pleasure to have with me Ruth Harpson CEO for Fairbanks One of the key partners in Netherlands. They are open a stack ambassadors in Netherlands They have been working with us and all this way providing feedback and validation of what we're doing So I will introduce you in a while before I want to introduce my colleague and also friend a big day our vice-president for engineering is just because you have been two years more than me in the Company, I wonder you covered a little bit about, you know, what's very test? What's the The BGN and what are we doing here? Okay? Thanks Carlos Yeah, Abhijeet some of you know me. I think most of you know me Been in this company for a very long time actually 18 months 18 years. Sorry 18 years. I wish 18 months Jet lag. It's jet lag. So You know we'll present the company, but you know today Veritas has emerged from you know We used to be acquired by Symantec one year back. We got freedom We are run by world's biggest equity investor Carline investment In a startup mode, it's great for us who built the original startup for Veritas to kind of come out Into a stealth mode startup now and it gives us opportunity to again Remerge as the storage and data protection leader that we used to be right. So today the company is organized into really three broad sections data Management or storage management that we proudly do for the last 25 years We did the first file system volume manager the industry had seen over the next the second portfolio is the data protection or Backup backup is our flagship product today. We call it net backup And then the third portfolio is the data insight data insight Ranges across the solutions that we have around Data governance data insight looking at data visibility of the data. That's really what Veritas is It's very simple data management data protection and data governance, right? So What what the heck happened? You had to click on the right bottle. That's okay cool So quickly on our history right because it's it's from the past. You know how to learn the future We have been you know, you know wedding in storage management for a long time First volume manager Unix world had seen first file system truly politics based file system Which we still ship called cluster file system and then the first ha or vcs solution The industry had seen back in the days 1999 2000 moving on to that We are adopting and emerging and reshaping ourselves to do multi cloud customers are going to go to multi cloud No matter what that's the next world We are going to see for many years and then software defined storage software different storage is a term industry Used last four or five years, but we actually defined software defined before it was a cool term really If you look at volume manager file system, this is a way of how do you? Provision storage managed storage in a software defined way. So we are emerging there, right? That's really where we are coming into open stack Open stack is a great Venture it's a in from IAS It's a great infrastructure for us to add values in storage and backup the things that we do and we do better than many people right many other Vendors right and we are trying to add value. This is a great opportunity for us to really add value in a more open Architecture agnostic way. We are checking in all the sources to open source. So this is a great idea Personally open stack. This is my I don't know 9th or 10th whatever number of open stack has happened It's it's almost like a festival Religions thing for me to show up and talk in this So I'm extremely honored to be now with Carlos and rude to talk about and launch the product on stage in Open stack summit Boston, right? So with that, I'll let rude talk about his company. Thanks Okay, Abhijit. Thank you. Good afternoon ladies and gentlemen. My name is Ruth. I'm son I'm the CEO of Fairbanks. We are a Dutch IT company And we are in the open source business now since 2011 and the reason we choose for open source is because it gives customers ability to innovate give them flexibility and scalability and also cost savings and We see a lot of examples like Uber and Airbnb What can give them a competitive advantage by innovative and flexible IT solutions? And that's the reason we are in this business In the last couple of years we built a lot of clouds for our customers and we did it with several distributions But now we have a very very good relationship with canonical and we build our clouds with Ubuntu and canonical and in 2014 I received a phone call from a friend of mine within Veritas and He phoned me and he said root Before I gonna say anything you have to sign an NDA and at that time I thought wow if somebody calls me in the evening and Wants to explain me something and I have to sign an NDA. It should be important So I received the NDA. I signed it in a couple of days later. I went in a meeting with with Veritas and I am really very proud that we as Fairbanks have The opportunity to help Veritas develop Hyperscale as where it is right now and yesterday it was the launch of a real very very good product To work in the open-source Community you have to add value and we do that in several ways. We have partners some partners in Europe Who don't have the expertise of OpenStack and we help them Developing and improving ICT and marketing concepts to bring their business forward We also have customers Where we act as a cloud specialist and broker of modular and open-source cloud solutions And of course we have our end users who we help and build our clouds the target market for For Fairbanks we have service providers and end users and as I said before you can see the picture for our server providers our customers We developed a service extender Which means that above on his services we can help them to improve their business and of course our end users Where we build an internal of external private cloud for the customers in the Netherlands and also in Europe Well, thank you. I think it's rich integration and again Thank you very much for being with us the reason all these all this time So we take a look to what problems are we solving for OpenStack. We concentrate on three key areas the first one is that We had to change the mindset about thinking about the story There's a whole of a story it's a volume that you go and use and you have to think more about the workloads So which are the specific needs you have to the workloads you can't have things like my SQL You can have things like Cassandra. Do they need the same requirements? So typically what you find is that this one side fits all so you and having maybe three copies for everything Do you need three copies for your analytic work role? Which is running a memory and you have three copy three copies of memory and then having another Three copies spread three copies spread three copies and it's just about the number of copies It's about how you are using the IO Which is go to the next point is that how do you really get the performance that you need? And definitely there is some progress and I'm working on the community about quality of service Okay, but how you can guarantee the quality of service that is Defined for your specific workload and make sure that the end you get the mass of your hardware Because you can be using commodity hardware, but how do you make sure you get the base of that? And I think here we're going to be talking about the architecture and the mindset So a big eat introduce a little bit of the history three years ago We started this project called open flame that today's hyper scale, okay? And that hyper scale was mindset change about we have to do something different in how we provide this and this is that One of the key things is that change the mindset about the workload because that is what really matters Okay, and when we think about that the third point which is Not very well solved or unsolved is how do you protect those clothes workloads? What is the protection we as veritas as the backup company? It's key for us and we we saw the key notes that more customers are moving workloads into production That's clear because more customers are asking us about how do I protect my workload if you need protection That means it matters for you So to solve this we are introducing hyper scale for up in a stack The first distribution that we announced yesterday is based on canonical I really appreciate canonical all the help to to go to a first in our NDA is to be agnostic and Of course, we're working with other vendors. We're working very closely with red hat and helping a lot We are planning to be with them by the end of the year and we have to say we have me mantis It's our NDA. So I did talk about so we define a storage We did that agnostic to any hardware to any story turn network to any cloud And of course, we have to be agnostic to any Technology that you guys may want to use that's in our NDA So what we really provide with hyper scale for up in a stack? So hyper scale for up in a stack is so we define a storage that this going to allow you to define Which is the resiliency level that you need for your workloads? Maybe you need three copies. Maybe you need to support two compute failures one compute failures That's something you're going to define at the workload level The second point is that how you can get the predictable performance with quality of service There is a turn of very well-known in the market, which is a noisy neighbor Of course, we're solving the noisy neighbor But not only that is that we can guarantee a minimum Ios per second So we're going to see a little bit our architecture how it works to make sure that at the end Your workloads is going to get the performance that I need because it's a cloud environment So I don't know what instances are going to be running and where they are going to be running So I had to make sure I create mechanisms within my storage So that my workloads around whatever they want, but they are going to get the performance that they need And the third point is the backup Integration or not backup I call data management without impact Well, what I see is an architecture that really separates the performance for the data management And it's going to help you to really do a what we call zero backup window or run a backup whatever you want, right? so Before jumping into a big it It would take a look to what we can compare with that traditional So we define a storage architectures that are based on a single layer Okay, is that where you have your hyperconverse the servers the storage the memory, right? And we have been there. I mean again, we create our file system 1995 1924 we have cluster files then the last version for cluster files is then that as you guys I love Support after 128 nodes using commodity hardware a local storage But the problem with that is that everything is on the same place So you have this noisy neighbor problem where all the instances are going to be running there And you have to make sure you fix that because you don't know what they're going to be running The second issue is that everything is on the same plane what I mean with that is The copy that I need now and the copy I need for resiliency are happening on the same place and if I go to Typical things like one side fits all where you get three copies for everything It's not about the number of three copies as you and I was talking yesterday Is that the number of Ios that I have to do because I make a right and then I make nine Rights that what I do those nine rights on the same layer. So everything is there So we took a step back. I think it was three years ago and say we had to rethink how we make this thing Okay, so I think that was a challenge for an engineering that was really well presented and a bit I'm going to let you cover sure what you guys have to do. Thanks Carlos and you know interestingly what you said It appeared to me again and again that if you have multiple copies It's not a bad thing right necessarily because you need copies for resiliency and ha purpose dr Purpose so you need copies, but point is when you're going to hyperconverged or cloud computing, right? You are mostly going to be using flash for your local compute storage Now we have these copies of the data so many times in the flash storage And that's going to be also very expensive because you're using two-thirds of your data center on flash storage Which are just simply copying the primary data and that's where we wanted to rethink right and so you know and thanks Carlos That is a great introduction. So if you really look at it We thought about this for 20 years and we challenged ourselves to say well for cloud architecture None of this storage architecture, which we have today in a legacy and also many other vendors today They will scale because you are stuck with one plane Which is called the compute plane where applications are running like Oracle or couch DB and where you take the backup Right data really has two layers, right? One is the primary consumption which is applications run. That's the primary usage. We call it primary data The data has the secondary consumption which is backup ETL archiving. These are secondary storage The problem with our architecture traditionally has been we morphed this secondary and primary data into one plane So what we did was really simple now that we have it and we are under NDA We introduced another data plane we call it a data plane as you can see the top is the Compute plane which is a scale out compute plane which serves up the application IOPS and the lower one is the data plane Which is here right here, which is also Storing the data right but what we do here is by doing that like Carlos says quickly We can now look at solving these problems of quality of storage because now we have one compute to worry about One plane to solve where multi tenant applications are happening We can look at optimizing the storage consumption because we can push off all those extra copies to a load here secondary storage data plane we can look at how do we do higher density storage and Finally, how do we do data protection because data protection has to be thought of now that we have two planes One is the primary compute and the secondary secondary data plane We can actually rethink the whole concept of data protection, but we can say you know what we can just take that Backup of the data from the secondary plane. We don't have to Impact the applications which are running in the primary plane to do backup. That's really the IP here Right This is very disruptive and we have protected this for almost two and a half years Except with partners and chosen customers that we have got a lot of feedback There are customers among you who have given us a lot of I mean I can see Sri and other people who have given us a Lot of feedback on how do you do better? We have definitely morphed and today we were ready to be GA. So we release the product. So I'll move on Yeah, is that so What are the key things is when talking about resiliency is that Don't know how many of you are familiar with the open stack flavor I've been talking to some people and maybe not familiar with that But that's how you define your policies your SLAs the tag of machine that you need that is where like in Amazon You say I want this number of CPUs and it this amount of memory Yeah, but what about the storage with the storage you today just defined I want 10 gigabytes What we do to have a real integration is extend on the flavors to say I need This resiliency factor. So this is the protection level I want to get and this is the number of copies I want to have because again my SQL is going to be different from Cassandra So why I can define different flavors Hyper SQL is going to define by default what we call Brahms and silver and gold very imaginative names, right? So but you can define your own flavors or you can add the Properties to the custom flavors that you want so you can create your Cassandra flavor if you want and define that an immediate need of this is the reduce on the hardware needs So we take a look to the economics and you take a look remember that single layer Right when I have my first copy my second copy my third copy my second and my third copy if I need a third copy Are you going to be we're going to move that down to the data plane? The data plane is going to be high resiliency and cheaper storage And the third thing is this is based on commodity hardware is so we define a store It's based on commodity hardware of your choice So is not this is my software solution and you have to buy my servers, right? No, this is really customized. So you choose whatever you want and we are going to be talking about the architectures a little bit. So Describe a little bit that how this works So this is the compute plane we have used this traditionally in any storage computing, right? This is the compute plane which runs your virtual machines or applications, right? So what we did this is open stack the instances are the Nova virtual machines which are running your applications workloads, right? There's nothing new here Now we have extended the flavors like our last mentioned about open stack flavors open stack gives you quality of service in our way Right, it gives you bronze gold silver which can be extended So we did that and then we said, you know when application rights are coming the primary The primary data is in this node, right, which actually gets the application and then we replicate The delta changes we do not copy the entire file over and over again into this peer compute That's what the traditional solutions would be that's how you end up with multiple copies All we do is we take the delta changes from the last 15 minutes or so Which is a tunable and we say let's only keep the last rights, right into the peer compute nodes That's the resiliency level, but then you'd ask them Well, you don't really have a full copy of the file anywhere, right? That's the next question That's where we said, you know what we can disrupt this we can say well You can you can use this data you can store this data from the primary compute plane Every episodic interval we call it episodic data sync But every episodic interval you can say well I can store this flush them to a secondary storage now The beauty of this design is once you are done flushing them into the secondary storage You can remove those deltas or replicas that you had in the primary compute so you end up with One primary copy in your flash tier and one secondary copy in your cheap and deep secondary storage tier That's really the solution Right now if you really look at it This is all works with open stack of course we had to upgrade the Cinder driver to look at local flash storage We had to upgrade the Nova to look at the filters that Nova provides to upgrade them to use this design So think of this like us to enter into it storage management for open stack, right now This is a storage technology, but we call it also end to end because now that you have the data in your secondary node Right, you can take a backup simply from that node. That's really the architecture So see the other thing to point out is that these four nodes we are showing four here These are scalable infinitely scalable because there is no strong clustering between these nodes They're not connected by a heartbeat network like other solutions even our legacy solutions had right there layer 3 network Each node comes and goes you can infinitely infinitely scale the compute node You can infinitely scale the data node. So your storage needs don't dominate your application grows faster application needs The storage does not have to grow that way. They can scale linearly, right? So I'll pass over to rude to talk about Thank you Well, you heard something about resiliency and quality of service, but one of the first slides of This presentation you also saw something about performance and this is a real case where we were dealing with a few years ago This is a Dutch supermarket. It's called the spar and they came to us and said Fairbanks Can you help us building a cloud where we can have Iops and a lot of Iops because they build an application on an Oracle database and What they what they do is and that's their unique selling point There is a guy who is working in the supermarket at the time. There is a product. He it's the the the shelf is empty He pushed the button the ghost information To the warehouse there is a guy who picks the products put it into a container container into the lorry lorry to The store and they build it very sufficient very high-qualified, but If you look about for elastic clouds, then you definitely need Performance so we did a lot of testing and at the end we had to decide and not to use Chef because Chef was not able to solve this issue and they had to go back to the standard solution like send technology and Because of the the stress touched we We did for this customer you saw with the elastic cloud Necessary they were looking for it was not possible to use F and we really believe that together with the product like hyperscale We definitely can solve solutions for our customers in the market I will Thank you, and we will make sure we get that so that goes to the predictable performance with quality of service The way I see that and and we have seen different solutions all on the market is that Do you have quality of service that you can define for each of the flavors or each of the workloads again? how I can have Critical database running today in open a stack and I can make sure I can run other workloads like web servers application servers and everything is going to get that so using my Artistical skills, I try to represent that with different regions where you can go and share and say this is I have this workload That need this high performance. I have these other workloads that need a different performance One of the things we do by default is to avoid the noisy neighbor So that means is that now any? Instance is going to keep under control and it's going to is not going to make any noise That is going to affect the performance of other ones and one of the key points is accelerate performance So what we do is that we keep that first copy in the compute plane, which is running where your workload is there So the CPU and the story chart together We also use SSDs to our flash to accelerate Performance to accelerate rights and do that protection that the big it was talking about and make sure that that right is protected But it's protected in a very effective way So let's see the how part Carlos gets to tell me what to do and I get to do the how part of it with my team some of you are here So did it well so basically So basically what again just listen to me with the slides cannot explain This is a pretty complicated algorithm, but open stack actually you have to mean is to say the algorithm Yes, yes open stack introduces quality of service by adding flavors to the virtual machine when you create or Instantiate a virtual machine. These are called gold silver bronze with the each tier you get a performance guarantee But then open stack does not really solve the storage problem as an IAS provider It's not open stacks job. It's vendors job to look at it, but open stacks tries to solve with QoS or quality of service at the source level at the desk at the source level But the virtual machine issues the IO what really needs to be done is to solve the storage Quality of service at the devices in we call it the destination what the data has been written. So what we do is we simply assign the Application a sort of credit system. We say how prioritize is the application? Is it my Oracle or is it a noisy neighbor? The noisy neighbor is not prioritized the Oracle gets prioritized We take that weights or credits we call it and we transfer that Understand in the system then we go back to the device and we understand the device latencies for that application Then we map that to and then we say who gets to write if the Oracle came with a very high Max max IOPS and min IOPS a higher priority with a gold level We let the Oracle flow in and the noisy neighbor automatically does not get the share of the storage It's deep prioritized. That's really the quality of service now This is something of course happens within a physical node You have a multi-tenant physical node you can say well I'll assign priorities and do quality of service between the workloads That's how we do quality of service our vision is to also take that and Map that into an entire compute plane where you can do quality of service across physical nodes That's coming up later in the road map So with that I'll let Carlos or do talk about Carlos so this is quality of service on on action and I understand it's a very complex It's like that. Let me go through the key things This is think about your database running where you have defined your flavors You want a minimum is per second up 10,000 and a maximum of 20,000 is per second you have some application servers between 5 and 10k and What happened here is that I start my database running it goes to the limit of 20k I have the application server goes to the 5k and then what I'm doing is a starting What I call noisy neighbors with no control what is happening is that? My database gets impacted because I have a number of ios per second I can share in my compute that's getting pat it but it's going to get just impacted to the minimum ios per second That's the the low band that I have on the 10k ios per second for my Database for my goal. So at the end what I can do is to have a better Share of resources within my system But also what I have is the capability to at one point in time go and limit Say now you are web servers are not going to get more than 1,000 ios per second And the moment I do that and I can do that online is that then my database and my Application server goes back to the violence that they had something I can do also is to recover that and say yeah now I have more bandwidth so the database is going to raise to 25k so I can give you 25k and Get the performance so this is an interesting Example because we're having 60 thousand ios per second in a compute plane And of course this is going to depend on what hardware you have. This is one SSD car on four HDDs That's all and this is a scaling in this testing We're running eight nodes having the same performance on each of those eight nodes So that's close to four hundred eighty k ios per second in an eight node cluster using commodity hardware, right? Thank you We discussed and picked this example out of our customers It's the name of the company. It's revised and it's a service provider And what we see in the area where we are working which is the Benelux Netherlands Belgium Luxembourg and the area around be a bit of UK and Germany. We see that those service providers They work still on a traditional way. They use their traditional software. They earn money with the customers the customers bring the VMs and on their platform and that's where they make money with and we believe that together with the cloud Software like OpenStack you really can bring more technology and more service for your customer So revised came to us and said Fairbanks. Can you help us building a cloud? We are really Looking for reduce of costs and we don't want to have a vendor lock in But the quality of services for a service provider could be and maybe should be very interesting for their customers So he has his as a lay to his customers and with a different quality of services. He can help his customer So now with this product like hyperscale Well, we are We can't wait. I'm a cheat. We talked several times now We can't wait to bring this to the market because we really believe it brings something for our customers like The service providers, but also the large enterprises good Thank you, Ruth and I think lastly but not less important or even more important is the data protection part Okay, if you have understood the explanation from a big it and I hope everybody understood a big it quite well You don't have a Boston accent like me, but that's fine You have this port in time copy, which is happening at the data playing every 15 minutes That's my first line of defense if you want those copies are going to be in your timeline your view of horizon So in your UI, you're gonna see how many copies you have you can go anytime click on any of them and do a restore More important is that we're using that data plane and the copies we have at the data plane to provide a zero backup window. I Hope you understand right now. You can run a backup. Whatever you want that backup is going to go to the data plane And of course what in what integrating with net backup and they're gonna show you a very quick demo about how it works Yes with one click you can define. What do you want to back up when and how? Yeah, so thanks Carlos so so basically it's now that we are gone through many times, but we'll reinforce right This is the data plane you're looking at you are looking at the point in time copies That's coming from the compute node right on the on the left arrow there So the it's a version storage, right? So you got the original data of the file could be a VM DK and then you got the incremental snapshots 15 minute Delta's that are now accumulated in the data node now It's a matter of taking that data from this data node into any backup vendor in this case our flagship product Net backup so now you got the data you just over API right You send it to a backup device and then the beauty is now you can restore it back and once you restore it back It's a one-click restore from the UI Horizon once you restore it back now you can rehydrate it back We call it rehydration which is meaning taking the data from the data plane back to the compute plane now You can reinstate instantiate rerun your application based on that data that you are restoring it from that's really the end-to-end Open stack net backup or open stack backup. That's something that not many vendors have been able to do But because of this disruptive design we have now not only solved an open stack cloud backup But also more importantly is we are not going to impact the application the data is already here So we don't have to go back and take an application whizing all the time for crash consistency This is perfect for application consistency You still have to say backup now from the compute node the data still shows up in the data node And then it's a matter of taking the data and tear it to your backup device You could even take it to the cloud right? That's kind of how the data protection portion of the solution works Thank you Another example And that's what what you said abhijit and also you Carlos I really believe and I see a lot of examples in the market Enterprises but also universities and Other companies they really have a problem if you talk about private cloud and then what can you do with your backup? And you said that maybe Not many vendors. I think you are the only one who really has Solution for backup for instance for this customer like the University of Luxembourg What they do is they work with several universities all over the world and do research on very difficult diseases And bring information to all the other to all the other universities And every time when I speak with the CIO together with my colleagues We have visited this customer. They ask how can we solve this issue because it's very very High valuable you can't Put a price on it because that information is so so valuable for those organizations That they really want to keep it for themselves with all the other with all the other universities But they don't want to lose the information. So they're really looking for a solution like what you have right now I'm pretty sure about that. Yeah, so let me show you a quick and demo how This really works and is this is out of the box integration with net backup The way it works is that now this is your horizon UI. You can see hyperscale tap You are going to see backups so in the net backup site You just create a policy that policy is going to define when you back up with thing Open a stack you're going to define for this policy when this policy is happening This is what I want to do and you want to do is I want a backup based on a flavor or a flavor and a name or a flavor on The state so you did really define what do you want to back up then when that backup is coming That's going to the data plane the data plane moves the data to the net backup And then we think that you are you're going to have the possibility of what backups Do you have you can just go and click in one of the images you have do a restore That restore is going to talk to net backup It's going to bring the data from whatever it is back to the data plane And then from the data plane you are going to instantiate a new instance that you were having just with really One-click restore so again out of the box integration so I asked for my folks in engineering about hey, I want to see a proof of How this how this works, you know how that's not affect performance and it's a very silly thing because I have 25 instances running having this perform and I said okay now run a backup. Yeah, they have the same performance So I don't know if I make a graph. We just as flat line Okay, this will be a silly slide, but that is the result that is what really happens So when talking Summarizing a little bit about the value probes that we have and when we think about the economics of the solution Okay, is that you think about that single layer that high percentage in a traditional so we define a story solution And if you are ending having three copies everywhere, okay? The proposal with hyper-escalar architecture is that that first copy is going to be in your compute planes If you want to do an all flash all SSD and you have a lot of vendors here talking about flash I can keep the first copy in flash if you want that's what I need The second the third copy whatever you need it's going to be on your data plane That's going to be higher density and cheaper storage And not only that is that hopefully you realize that what's happening is that the IO paths are Much more better utilize so now I have the is West network what I have my resiliency my compute to make computer optimized for performance my data plane Is made for data management so at the end if you have the hardware you have the software you have all the components We really have we can have a very good proposal in our TCO model for our customers So thanks Carlos So if you really look at it summarily this is open stacks horizon console, right? What we have done as a storage vendor right who understands storage We have added a storage persona to open stacks horizon So when you have open stack installed you will see these dashboards which you don't see with open stack vanilla You can look at it. There's really nothing here about IP So we are actually showing the primary compute plane That's the primary compute on the top the date the secondary storage data plane is there You will see all your virtual machines You can drill down to the applications to your physical nodes You can see the performance characteristics Carlos talked about backup You can drill down which virtual machine into backup you can do all of that right from horizon So we have pinned down into horizon checked in the changes of course It's needed to open source contributing into the community which we want to do and then that's what you see when you install The product on open stack be it canonical or red hat or Mirantis or any other Distro because we are distro agnostic right we are cloud agnostic The other last point I want to mention is that this is a storage technology We are taking this Technology from open stack into containers like dockers and Kubernetes And we are already working on solutions because this storage technology is a cloud scalable hyperconvert software defined technology There's nothing really pinning down to open stack open stack is our lunch ground to get the product going in solutions Tested with you customers who are going into the open stack journey But we are going to take this technology moving to dockers and other cloud friendly ecosystem with that I'll move to Carlos. So please visit us in the booth We are just here. This is Bankia. We're director of engineering thinking about what's the next thing he's going to be building and With that we will finish with a video from our friends From the stack is it working? Hi there we are Fairbanks We are part of the open stack community a community of users and developers building the ubiquitous open source cloud computing platform The use of open-stack cloud software gives you the ability to innovate the get-up and running is very quick and efficient We help you to avoid a vendor lock-in with 24 7 production grade support save costs and you pay as you go Increase your strength with innovative cloud solutions We build a stable working open-stack cloud for your unique needs and provide you with training and support The open stack dot nl sandbox is the simplest way to try out open-stack cloud software Therefore we've built a continuously growing open-stack test environment with hardware that has been made available by several suppliers The use of the open stack dot nl sandbox is free Here you can test how your applications operate on an open-stack environment Sign up for the open stack dot nl sandbox Good, so that's great, and I think it's an imitation for Yes. Yes, so We built the sandbox just to give everyone the ability to See how open stack works, but also now with with hyperscale I think we did sign up when you arrived here in this To listen to this presentation So we will come back to you and give you the ability to Sign in to for the for the sandbox and there you can see how open stack together with hyperscale works for you Good So what's that? We really want to thank you all of you for being here I don't know we have time for a question or it's very late with any question you may have Any quick questions now and we can always meet at the booth tomorrow. Thank you guys. I think Nice solution. I think the QAS and T-Renkella kind of helps Some of the solution we've been challenged with at Verizon wireless was how do we provide a Instance local storage with low latency and high bandwidth That could be couch base or Cassandra or some of the databases that you have That requires Lowest latency possible so that she can you can do a lot more throughput across the instances that you have the cluster that you built So I think what the hyperscale kind of we're trying to figure out trying to work with it so far is to Provide the ability for the instance to get the local storage and have that affinity defined so that the instance will have Local storage by the high performance range when I say local storage don't mistake me for just local storage We want to put SSDs or in VMEs or maybe flashcards to get that high performance through it But coming from the local storage to the local instance so that you can build that cluster and say, okay the instance you have your high performance storage locally available to you and Go crank it up your database as fast as you can. So I think that's the kind of use case We're trying to do and I think thank you guys Abhijit and relation all these guys to kind of get some POC going on. So just want to let you know We're almost there Thank you. Thank you. Good to hear from Fairbanks also that you guys are already putting into production. Thank you I'm just seeing the use cases already. That's good. So I think we just get a lot more cranking up to do but in the meantime just Good to be good Thank you. We really appreciate you coming along helping us ship the product and again again in the open source Community, this is like fun for us engineers, right? You can see the source code of other vendors who knew and then you can contribute, right? That's that's really what we are doing again Of course, there's a business aspect of it But it's glad to be here among the community and today's ring of really Historic day for us to be doing something for open stack actually contributing into the open source And then helping customers as they move to their cloud journey. Okay, but yeah, personally It's a dream come true for me. Honestly, but Carlos great moment. Thank you everybody. Thank you for sharing the moment You're here. I see how many times 10 p.m. I trying local time your face on a web Exhibit wake so you guys are there. Thank you for all the work and thank you for coming It's pretty late in the day. We understand. Thank you