Aaron Sullivan, Rackspace | OpenPOWER Summit 2016
hi this is David flora back at the open power foundation conference here in San Jose and with me I've got Erin Sullivan who is a distinguished engineer at Rackspace welcome our thank you so what do you think of the conference so far it's amazing it's grown so much in the last year 15 designs to almost 60 in a year and lots of system launches yeah very impressed well one of the things that has been announced today which was caught my eye in a big big way was the agreement so the announcement that you and Google have can you paint a little more put a little more light on that announcement yeah sure so Rackspace and Google started working together when Rackspace was developing barrel I of course Google had already had their system available at the time and our collaboration just on what we had with barrel I was very positive we were just kind of looking to trade notes and you know share our experience and a few months ago we got back in touch and said hey this was ministers posit enough we should think about doing the next one together from the start and so that's basically what we're doing now we're going to do a power 9 system that comes in multiple mechanical form factors but just one motherboard and we're going to like we did with barrel I we're going to contribute that to open compute when we're finished out of the Open Compute foundation part of the OpenStack yes heart of the open power founder that's right open everything open ever yeah yeah excellent so what about the barreleye that you also announced some things about today can you what is barrel i and what's what's what's different about it so so paralyzed named after a fish that's got a transparent body most of our servers are named after we thought having a server that was fully open would be great to have that name barrel I just entered its first data center shipments it's headed to our Virginia data centers right now and in a few months we expect we will begin providing services to customers on it so that's the progress on barrel I so far we contributed to open compute about 2-3 months ago now and it was accepted so the specifications are online and if you look around the show floor here you will see there are other companies that have put their brand on it or something else and are also taking at the market which is exactly what we hope for great well I've got a question which is why have you why have you put these resources into barrel I and in the future into the power 9 etc what are you looking for that's different about open power that for example you couldn't get with a standard x86 server yeah so I know it gets to be tired and people get tired of hearing the word open but really even with open compute and OpenStack the freedom that comes with developing in that particular universe is really significant before open power even started there were parts of the system we really wish we could get into in an open way where we could develop and share instead of just doing it all on our own and having open power come in the first place fit that but then we also have this problem this Moore's Law problem and the types of changes that we're going to have to implement as an industry to continue to accelerate and and and get higher performance computing and more efficient computing over the next year's they're really huge challenges they go from the chips all the way to the top of the stack and if you don't have the chip part open and you don't have the firmware part open it becomes really difficult to collaborate you can't bring to bear the sort of force of the world software developers onto it you end up in these little silos and niches so for us beryl I provides a lot of value as a business and it has a great influence on the industry at large and so wills IOUs the power 9 system Google but it also is there as a platform for developers to begin to start wrapping their minds around these new problems and opportunities that we have and if it's not done in the open these types of software aren't really scalable across the whole industry that that's a very interesting answer indeed and as you say um does laura has come to a screeching halt from the point of Mount of power per CPU is still going on in terms of the number of transistors etc that you can have what are the what are the things you as a distinguished engineer what are the things that really are most important about the power architecture that allow you to develop these new ways of doing things yeah I think it's it depends on the type of your business you're in but in our business I think in many cloud service providers and in some other environments certainly some HPC and a lot of enterprise the performance of a single core is still really important and it will continue to be for as long as we can keep getting more performance out of a single core so power provided a great platform with a very powerful core and it also has a huge number of threads per core so you get a little bit of the best of both worlds there and you need a really powerful core you have it if you want to spread your load really wide over a more cloudy webby type application you get to use all those threads and there's all that memory bandwidth and so forth so so that was the benefit of power in general and then we run out of core performance and those cycles per you know CPU aren't going up and maybe we can't even scale cores like we used to anymore which is coming in a few years I the the fact that the platform is open in areas that others aren't allows us to bend the rules about how components communicate and we cut out a lot of overhead between them so that's a sort of software in silicon type argument you want to bring the software closer to the silicon yeah closer and in many cases to do the same work that we do today like that's the hard part is people think it's all about genomics or oil and gas or something it's the same work but you know we've already demonstrated that open harcum you it is demonstrated that there are certain workloads that are very common today that you can boost tenfold or more simply by reintegrating your software tighter the hardware right you pull out overhead that we were fine with when Moore's Law is working but now we got to do something yeah great well thanks very much indeed for for being here and thanks very much for watching
**Summary and Sentiment Analysis are not been shown because of improper transcript**
ENTITIES
Entity | Category | Confidence |
---|---|---|
Erin Sullivan | PERSON | 0.99+ |
Rackspace | ORGANIZATION | 0.99+ |
San Jose | LOCATION | 0.99+ |
ORGANIZATION | 0.99+ | |
Aaron Sullivan | PERSON | 0.99+ |
Virginia | LOCATION | 0.99+ |
last year | DATE | 0.97+ |
today | DATE | 0.97+ |
David flora | PERSON | 0.97+ |
both worlds | QUANTITY | 0.96+ |
next year | DATE | 0.96+ |
single core | QUANTITY | 0.95+ |
first | QUANTITY | 0.95+ |
laura | PERSON | 0.93+ |
single core | QUANTITY | 0.91+ |
one | QUANTITY | 0.89+ |
first data center | QUANTITY | 0.88+ |
2-3 months ago | DATE | 0.84+ |
OpenStack | ORGANIZATION | 0.84+ |
a few months ago | DATE | 0.83+ |
OpenPOWER Summit 2016 | EVENT | 0.82+ |
almost 60 in a year | QUANTITY | 0.79+ |
OpenStack | TITLE | 0.75+ |
one motherboard | QUANTITY | 0.72+ |
15 designs | QUANTITY | 0.7+ |
about | DATE | 0.68+ |
x86 | OTHER | 0.68+ |
few months | QUANTITY | 0.66+ |
power foundation | EVENT | 0.61+ |
tenfold | QUANTITY | 0.61+ |
few years | QUANTITY | 0.58+ |
lots of | QUANTITY | 0.54+ |
things | QUANTITY | 0.52+ |
Moore's | TITLE | 0.48+ |
Moore | ORGANIZATION | 0.33+ |