 Good afternoon everyone. Thank you for coming out this afternoon for those of you in Tokyo Thank you for staying in your country and those of us from outside. Thanks for making the trip to be here with us My name is Anne Fenei. I am the director of power systems enablement I have the great fortune to lead a wonderful team of developers at IBM that essentially builds the site software stack the entire Software stack from our hardware all the way up to our open stack solution power VC Many of them who are here, and I hope you'll get a great chance to interact with So I want to start talking to you a little bit today about a new era of innovation if you think about everything that has happened over the last I mean 50 hundred years how technology has changed from things like putting a man on the moon to just the release of the personal computer Which an interesting fact there the first personal computer came out in 1983 Had a 10 megabit hard drive and cost about $5,000 Quite a change and quite a lot of innovation has happened since that time and today now we talk about the internet of things we talk about Everyone being connected we talk about every device being connected and constant connectivity all times everywhere anytime Things have really changed and evolve rapidly Now start thinking about how you drive that innovation and how things have changed Innovation initially started You'd have one company or one person Now you see it driven by communities of people partnerships of companies and there's been this Realization or this coming together that no one person no one company can do it alone And that's why groups like open stack are so very very important Talk about the Linux Foundation these organizations are so important because they bring diverse thoughts diverse experiences together And really drive innovation take it further than it ever could if you just had one person or one group working on this stand alone At IBM one of the things we really believe very very strongly in is creating open technology innovators Not just from my area of the power systems group but from other areas within IBM as well There's a huge IBM presence you know we believe at IBM of participation give back to these communities And we want to through things like the academic initiative we want to create that next generation of open technology innovators And because I am part of the power brand I do have to share with you one thing called the open power foundation And this is something we're very very proud of it started about two years ago There were five initial members IBM Google Nvidia Melanox and tie in Together those five companies said we want to provide vendor choice for hardware and systems for clients everywhere To provide opportunities besides the names that we all know that are in the marketplace today that have large market share We said this is really important. Can we provide chip architectures open to the community? My team in particular we've provided our boot firmware to the community our on chip power and thermal code controls to the open community These are very very important and it's really been a commitment from IBM systems as a whole that open is not just the top of the stack The open stacks the Linux it really is going to be at our core This is so important to us as we go forward that open means our systems all the way through our software So with that I do have to give a little bit of a marketing pitch to everybody in the room and I hope you'll appreciate this Because it would not be without this open power foundation that we could announce our power systems LC line This was really a revolutionary Line launch for us in power first It is the first ever power system that you can go buy online with a credit card at IBM You can go to IBM comm slash power by one of these machines They were built in partnership with our open power partners You'll see in the lower left that you can actually buy a system Optimized and integrated with Nvidia's GPUs one of our open power foundation partners We've optimized for specific workloads that are very important to hyperscale computing to cloud computing And we've made sure throughout our power systems line as well again cloud computing open stack summit I'll give that plug as well We want to make sure we are very we participate heavily in the community and we give back to this community Because we believe these systems are really going to be so important for us for the future So with that I'd like to take a moment and introduce a great client of ours and partner Clayton Weiss from key information systems Who's going to tell you about his journey with IBM and power systems. Thanks So I'm Clayton Weiss I am the director of cloud services at key information systems We are a regional cloud service provider and internet service provider based in Southern California About 30 miles north of Los Angeles I've been in the service provider industry for the last 15 years or so And been in technology since about as long as I can remember So key info acquired the company I used to work for And so they decided hey there's something in this cloud business that might catch on We should probably get involved So I want to say first thank you for everyone for for coming here I really appreciate the the time that everyone's putting into this and our partnership with IBM Alright so first is I want to start with a slide Actually the slides a joke there's really no valid information on here I was just told that the thing you should never do is start with a slide So I had a client come to me about five or six years ago And he said all these data centers all of you guys you're all going to go away You're all going to go out of business and I said why And he said well everything's moving to the cloud And I said where do you think the cloud is All the cloud is running inside of data centers And so it's important for us to remember And we actually believe that in this age of growth And in the building of these different clouds that are out there The regional cloud provider becomes an integral part of how most of the The mid market and the larger clients are going to consume the infrastructure For some of the born on the web and those types of new applications Maybe not as much but for a lot of traditional businesses And for a lot of companies that need that There's absolutely a need for a regional cloud provider And for regional providers across the landscape So key info and IBM have a really long partnership It goes back almost 20 years And one of the things that I had never dealt with IBM two years ago To me IBM was a company with three letters and they were blue And that was about the extent of my knowledge and experience And since then I've had a lot of exposure to the type of systems And the technologies available for my IBM and I got to say I'm kind of sold We actually recently at the Southern California Linux Expo Did some testing of Intel versus Power Systems To do a bake off and see how applications performed in them We did one test for example We generated a million prime numbers on power And a million prime numbers on Intel And this was the latest generation Intel chip 3.6 GHz fastest memory available It took 238 seconds We did the exact same test on the power system It took 22.3 seconds We've done some bench testing of MariaDB and Postgres And we see anywhere from 2 to 3x the performance difference So for really powerful workloads The power chip provides a big advantage The other for us is these systems are built a little bit different Than the standard sort of disposable white box That you might buy from like a super micro Or to some extent Dell and some other different people These systems come from a heritage in the IBM world When servers are built to survive If the main frame went down The entire company went down with it These systems are built with that in mind On the higher inversions of these servers You can actually replace IO cards And add new processors on the fly Without having to power down without having to reboot This becomes important when you talk about How do we migrate enterprise workloads And get them to adopt a cloud environment Having this type of infrastructure becomes very important to them Because they haven't ported their applications over yet They haven't moved to that As an idea of how bulletproof these servers are One of our techs when installing a system This was a $40,000 AIX server He dropped it from 30U up And it went face first right into the data center Punched a hole in the tile about that big No damage to the server So that's one of the reasons why We really like working with these systems Is they're extremely reliable and they perform very well So in the world of cloud You have two choices And how to build your infrastructure It's very important to understand That how your application is designed Your infrastructure should match So these days When we consume everything through an API There was no thought to the underlying infrastructure This used to be a discussion 10 years ago That I used to have of matching clock speed With memory speed to make sure that we had Optimum efficiency in a server These discussions don't happen anymore And it's not to say that That's a good or a bad thing But what it means is We have a tendency to forget how important it is That the infrastructure design matches the application A lot of traditional enterprise applications Are not built to scale horizontally If my MongoDB cluster is running slow I add 10 more nodes But if my AS 400 is running slow I don't add another one I scale up The infrastructure from the provider needs to match How your application is designed And this is one of the big challenges That a lot of our enterprise clients have In adopting cloud They'll find themselves spending $20,000, $50,000 A month to get around performance issues That if the infrastructure was built to scale The way that they were Save them a lot of money and make a lot more sense And get much more predictable performance So for us, we need to be able to accommodate Both types of workloads within our cloud environment This is one of the other challenges In OpenStack, it was a joint project Between NASA and Rackspace To NASA, all workloads are cattle To the enterprise, all workloads are pets Trying to explain to an enterprise That when you boot from a glance image And shut down your system That's ephemeral storage, it goes away It's a very difficult conversation They don't like the fact that When they turn off a server, it disappears forever And many of them, due to the way Their applications are designed Are not built to put things on separate drives And have interoperability between systems in that way Everything was designed to scale vertically Every virtual machine, every system is a pet One of the biggest challenges with OpenStack adoption in the enterprise For us as the reeducation of those clients It's training them on how to provision And manage their systems in an OpenStack environment That provides them the ability to scale on demand The ability to grow, but not to lose The functionality they're used to So we usually spend about the first Two to three months after bringing a client on Just going through training and getting them up to speed On how to truly manage their infrastructure In a cloud environment Networking in OpenStack has been One of our biggest challenges Prior to Havana and really to the Widespread adoption of Neutron Networking was an afterthought It was one of those things within OpenStack That it was a couple to the virtual machine It was just another... It was like metadata, it was some other piece of it It didn't stand on its own In the networking world And I come from a networking background On the service provider side The thought that network is something That's just bolted onto a VM is a problem And it's a problem because When the enterprise wants to adopt cloud And make use of it They don't want to give up control And in the past Networking was one giant flat network And it was assumed that everything was public facing Many of our clients, everything is private first Public second, and in some cases Public not at all So this is one of the big challenges This is actually a key area That regional providers can help solve for When we talk about hybrid cloud How to make use of that and how to Whether it's to cloud burst or to push workloads Out to different clouds for diversity Or due to opportunistic reasons Networking is going to be a key component of that This actually is an area that has yet to be completely solved We are a big supporter of Open Daylight And we like what it's trying to do And creating an open standard as a way To inter-operate between things Although it's not quite there yet We are working on different ways to integrate Our client networks into what we're doing The ultimate goal is that When a client moves a workload to us They can still control the network and the governance From their site Right now we can at least provide them With the same enterprise feature set And the same type of things they are used to doing Through a little bit of trickery on the back end Of allowing them to continue to control The network rules and the associations By doing some custom networking on a client-by-client basis They can still manage it through a neutron But it takes a little bit of work in getting it set up But this is an area that we believe Regional cloud providers can help Because distance matters When we talk about the speed of light And fiber optics Having an instance that is 30 miles away Is very different than 300 And when the enterprise goes To adopt a cloud environment Especially when you talk about applications That are not geographically aware It becomes very important to have that So we actually have a fiber network In the area that we serve That we're able to leverage for a lot of our clients And that is something for us That Amazon, Google, Rackspace, Softlayer None of them can touch It gives us a unique ability To provide an environment That they can't get anywhere else So in the cloud world Many of us would like to think That our neighbors are good and friendly And they're not So for us as a cloud provider We have to figure out how to deal With the noisy neighbor problem The way in the past that Amazon And some of the others have dealt with this Is by simply telling you provision more instances That doesn't really work for a lot of our clients Because, as I mentioned before Their applications don't scale that way So we have to find ways to deal with this So the key to this is Looking at ways that we can troubleshoot We use a number of special tools And things that we do to help That you can kind of only do In a vertically scaling infrastructure To see where your performance issues Are coming from and to help solve for it In addition to that We do something a little bit more brute force We have a lot of flash To give you an idea of some of the things That we can do with flash We're in the process of discussions With Redis right now To build out an offering There is a way with the IBM power systems And the special interface they have That lets us take our flash And present it as RAM to the system So the type of flash available from IBM Gives us microsecond latency Much faster than just about anything else on the market What this means is we can effectively replace An entire rack of 20 to 30 Redis servers with one server And two U of flash So flash is a very important component of what we do We have clients that will burst On a single virtual machine Up to 200,000 IOPS And this flash allows us to accommodate workloads Upwards of 400,000 IOPS on a per VM basis So we get a lot of really good performance And we try and make most efficient use of it One of the things that we're looking forward to That's in the current release But we haven't upgraded to yet Is the ability to enable compression Which will let us get even more performance per dollar Out of the flash infrastructure One of the things that we really love about OpenStack Is the diverse environments that allows us to provide So, like I mentioned We have a number of clients on traditional Unix systems We just internally actually led Some training for our techs On basic AS-400 operations So we want our internal first line of support staff To be able to provide some basic testing And troubleshooting level one support for the AS-400 And at the end of the training One of the techs raised a hand and asked If we could have an instance set up for them So they could go test and do some of the things they learned And one of our AS-400 operators said Sure, give me about a week I need half a day to go get everything set up In this, configure that Have the storage people Create some LUNs Have the network people And what to him normally took Half a day or more to do I turned to one of the cloud technicians and said Can you provision an instance of I inside of Power VC And in 10 minutes it was done This is something that In our enterprise world Our clients are not used to doing And it's what we've grown So accustomed to in a cloud environment of just click And it's provisioned For a lot of these legacy operating systems Is something they've never seen before OpenStack provides us with this ability to do that And in a way that allows interoperability Between them I can take AIX and IBMI And have it talk to a Linux virtual machine Running in VMware This is something that used to be very difficult to do And now is accessible through one API Talk to Nova, talk to Cinder Provision your instances All of these things can communicate with one another And so it's something that for us is of great value What it means from my perspective as the cloud provider Is I can change And nobody will notice We can switch from VMware to KVM To Zen To PowerVM To PowerKVM And none of our clients are ever going to notice They're going to provision and manage The exact same way that they always do And we get to change that out in the back end So we use a combination Of both IBM Cloud Manager And PowerVC To provide this layer of OpenStack On top of the infrastructure PowerVC is actually an OpenStack based product as well So in a way it's sort of OpenStack on OpenStack Which makes the configuration a little bit interesting at first So for someone that hasn't done it yet I will say it's vastly improved from what it was In its original versions But it does make for some interesting configurations at first PowerVC is a replacement of a tool From IBM called VM Control Which if you haven't used it Be glad you didn't PowerVC is very stable Works extremely well Is very reliable From the very first version of that application that we used We knew it was a really good product For what we were doing And so I have nothing but good things to say about it IBM Cloud Manager was an evolution Of a couple of different products And they've made a push in the new version To try and stay as close to trunk as possible And it shows We're very happy with the new version And the way that it provides the functionality to our clients Allows us to have the same Vertically scaling infrastructure on the back end With the traditional legacy systems For our clients running those Unix and IBM I workloads But provide it to them in such a way That they see it as a modern cloud control panel And so they have the ability to provision and manage And oversee their operations in a way That they've never had the ability to do before Any questions? Okay, well thank you very much So thank you again everybody for coming Thank you very much Clayton We do want to invite you to come down To the marketplace tomorrow to the IBM booth We actually will have a power VC demo going Again that's our open stack solution for power As well as the IBM Supervessel Cloud Which was done through the open power partnership We're very proud of that And also we have one of those power LC machines There for you to play with, put your hands on And see in person So thank you again, thank you Clayton