We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

Open Software deserves Open Hardware

00:00

Formal Metadata

Title
Open Software deserves Open Hardware
Subtitle
How Europe is taking the lead in Open Hardware
Title of Series
Number of Parts
561
Author
License
CC Attribution 2.0 Belgium:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Identifiers
Publisher
Release Date
Language

Content Metadata

Subject Area
Genre
Abstract
This proposed talk would focus on how Europe is taking the lead deploying and supporting Open Hardware technologies. Large enterprises are looking for best practices gleamed from hyper-scale companies like Facebook, Microsoft and Google who run highly efficient private and public clouds. The sharing of open hardware and data centre designs is a core strategy for these companies and the basis for the Open Compute Project (OCP). The Open Compute Project (OCP) was started by Facebook in 2011 with the idea of delivering the most efficient designs for scalable computing through an open source hardware community. We believe that openly sharing ideas, specifications, and other intellectual property is the key to maximizing innovation and reducing complexity in technology components. In this topical discussion we would discuss the recent research that highlights how efficiency plays a role in the data centre transformation from conventional gear to OCP and open source. We would provide specific examples of how local European VAR’s and Solution Providers are capitalising on the open hardware business model and how "Open Source Software deserves Open Hardware".
Open setSoftwareComputer hardwareTransportation theory (mathematics)Shift operatorVirtual realityVirtual machineHacker (term)Server (computing)Service (economics)Computing platformReduction of orderMobile appDeutscher FilmpreisStandard deviationType theoryOpen sourceProduct (business)Data modelBuildingPhysical systemCollaborationismScale (map)Modul <Datentyp>Power (physics)Data storage deviceFirmwareInformation securitySupercomputerExtreme programmingStandard deviationRegulator geneAreaSelf-organizationRight angleWeb 2.0Offenes KommunikationssystemOpen sourceComputer hardwareChaos (cosmogony)SoftwareArchaeological field surveyInformation technology consultingReduction of orderLogic gateFeedbackDigital signal processingYouTubeÜberlastkontrollePoint (geometry)Data centerTouch typingPresentation of a groupPower (physics)Open setEndliche ModelltheorieServer (computing)Service (economics)FacebookProjective planeVirtual machineScalar fieldData storage deviceMobile appShared memoryElectronic mailing listNumberHost Identity ProtocolSpeichermodellHand fanCorrelation and dependenceKey (cryptography)Denial-of-service attackReading (process)BitComputerFocus (optics)Different (Kate Ryan album)Covering spaceComputer animation
Type theoryPoint cloudCircleModul <Datentyp>CAN busConvex hullOpen sourceData storage deviceOpen setReduction of orderComputer hardwareData modelStrategy gameBuildingSoftwareFacebookSupersonic speedData recoveryInstallation artComputer networkIntegrated development environmentBucklingNetwork operating systemModule (mathematics)Commutative propertyNormed vector spaceFirmwareMachine visionBootingCodeArmMIDIIntelBefehlsprozessorStack (abstract data type)Computer hardwareOpen setBootingVirtualizationSoftwareOpen sourceStandard deviationCartesian coordinate systemFirmwareNumberFacebookProjective planeType theoryRenewal theoryOffenes KommunikationssystemAreaLiquidHard disk driveCASE <Informatik>Electronic mailing listPopulation densityTouch typingService (economics)Food energyDeutscher FilmpreisSupercomputerGroup actionServer (computing)Point cloudComputerGame theoryProduct (business)Self-organizationAbstractionSupersonic speedCombinational logicInstallation artAdditionMultiplicationSharewareCuboidScaling (geometry)Office suiteRight angleOpticsScalar fieldWebsiteComputer architectureInterface (computing)Integrated development environmentObservational studyCore dumpData centerDifferent (Kate Ryan album)Moore's lawJSON
Product (business)Server (computing)Maxima and minimaWorkloadSoftware developerComputer hardwareHypercubeIntegrated development environmentProduct (business)Server (computing)Type theorySoftwareAdditionECosStandard deviationOpen setComputer hardwareInternational Date Line2 (number)Different (Kate Ryan album)Right angleExpected valueEmailFood energyHookingSpacetimeMikroarchitekturCircleBitComputer animationProgram flowchart
Computer animation
Transcript: English(auto-generated)
Steve, who will talk about open software deserving open hardware. Let's start the countdown. All right. Thank you. Great. Can everybody hear me all right?
Wonderful. How many people are drinking already? Fantastic. My presentation is going to go great. Thanks again to everybody. Sylvester, thank you very much. To the folks at FOSDEM, let's get started. I'm going to focus on a couple of things. A little bit about software eating hardware. Cover the OCP Foundation and the projects that are taking place.
And then lastly, share just a couple of examples around how software is working with open hardware. To start off though, the most interesting headline I read yesterday in the news, how self-driving cars are going to be just cruising around to avoid paying for parking. And the article then goes on to say how regulators are then going to have
to come up with all new regulations on congestion tax and road tax. So I always find it interesting because my colleague always says that however fast open source innovates, regulators are always trying to chase them to do more. So keep up the good work and keep the regulators way,
way behind. The other thing, if you get nothing out of this, add this to your reading list, Architects of Intelligence book that just came out not too long ago. It's an interview style of around 25 or 30 of the guys that are driving AI right now. And many of them, you probably already know, many of them are brand new to me.
But I highly recommend it. All right, getting on with software eating hardware. This number is likely going to be a lot higher. So by 2020 or next year, Gartner's saying or IDC saying rather that 30% of this is going to shift over to open source. The customers that I'm speaking with now are probably in that 40% to 50%, if not higher.
Looking back prior to VMs, moving to 2008 when we came up with lightweight VMs and then moving on to 2018, you're seeing a massive reduction in the amount of hardware that's needed to run apps. And earlier, when Facebook created Hip Hop Virtual Machine
to cover their web servers, they were actually able to reduce the amount of hardware they needed by 80% or more. And just a couple of years ago, this right here is a YouTube video of the Deutsche Bank CTO talking about their digital transformation
and covering almost 2 thirds of their hardware estate gone, running 85% of their apps. So how is this translating into my world? Well, there is no doubt that Linux is pervasive across all of the data center infrastructure. The same is happening in ours. So we're seeing a linear correlation
between the amount of people that are adopting OCP and the interest that's happening within the overall community. The way our foundation works that's different than, say, a couple of the other organizations, in hardware, you have two extremes. You have standards bodies on one area.
It's where vendors get together and say, let's create a standard. You might get voluntary adoption from that, say, from one or two customers. Then on the right-hand side, you have multi-party sourcing agreements. That's where a bunch of customers get together and say, let's pool our buying power. And they may get one or two vendors to then supply that.
In our organization, you have vendors working with suppliers, working with end customers, working with consultants to all come together to pull out this hardware design. In the OEM model, you have the traditional Dells, HP, Ciscos. They go survey a bunch of customers,
or they get this feedback on designs. Then they may gatekeep some of that innovation based on the need that they need to serve 90% of the market. So in the open source model, there's this perceived chaos that happens between now I can actually work directly with the hardware manufacturer. So if I'm an open networking switch manufacturer,
I can actually go directly to that manufacturer and get my switch. I no longer have to go through a traditional OEM that wraps all their services and support on top of it. So moving on to the foundation itself, this is a news article from 2010 when Facebook first announced that they were going
to build their own data centers. So not only did they build their own data centers, they went directly to the manufacturer and said they started doing their own hardware design. But what was different about this compared to the other hyperscalers is when they got done designing their own hardware, they open sourced it. They open sourced their switches, their servers,
their storage, their rack, their power, and even their data center design. So in 2011, they started asking themselves, what can we remove? Can we actually run the servers at a higher temperature? Can we get rid of a lot of the stuff in the data center that we don't actually need? Open sourcing that into a technical non-profit foundation.
You had five other companies participate, Goldman, Rackspace, Intel, Microsoft. From that, you have the Open Compute Foundation project. 200 plus companies now focused on four key areas. Everything that we do that we bring into the foundation
has to meet and pass these four criteria. For example, all of our designs have a tool-less design. So you actually don't need tools to operate on the servers. You have touch points. A lot of the hardware comes with a larger fan, creating more efficiency.
You have shared power in the back. You have some flood cooling as opposed to big chillers on top of the data center. So a lot of efficiency designs that are taking place. The project itself, 10 right now. 10 working projects run by volunteers across all kinds of companies.
The most interesting, the networking is by far the most active. And then open systems firmware, which has just started recently, it's currently being run by one gentleman from Google and another gentleman from Microsoft. So what types of companies are participating? Don't get too bogged down in this giant list here.
I'm just going to pull out a couple, AT&T and Deutsche doing a lot around open Wi-Fi and open telco edge products. Alibaba doing a lot of advanced cooling for us. JD.com and InSpurr, so InSpurr in the top right, the number three server manufacturer in the world, open sourcing all of their designs.
There's some worldwide companies. Here in Europe, we have over 40 companies in Europe that are based and headquartered here that are participating. I'll pull out a couple that may be of interest. OVH, 2CRSI based in Strasbourg doing some really interesting
HPC oil and gas solutions. Rapid Space, I'll get to those guys in a minute. IT Renew. And then down here on the bottom corner, you see Submer. They're doing some really interesting liquid cooling for a lot of the large hyperscalers. So liquid cooling and advanced cooling technology has been the number one growing area of interest
in our community in the last six months. Because of the higher density racks, the more computing power, the traditional cooling methods are not gonna be able to handle it. On our website, you're gonna see a few things. Primarily, you'll see some of the older things here. You'll see specs, contributions, designs.
We've taken those for years. You'll see vast amounts of paper specs. Over on the other side, you'll see new things coming along, reference architectures, case studies, white papers, all coming from our members. Two examples of what's happening. Edge Core, one of our largest open networking companies, introduced 400 gig optical networking.
And then Microsoft, working on some new cloud hard drives. So these two projects specifically are born out of companies not being able to solve a problem themselves. What happens in the community is you have one company that raises a problem around hard drive APIs. They raise that question.
And inevitably, you have three or four different other companies that have the exact same problem and start to work on it. I pulled out these examples here because as you're working on the software piece, it's important to understand what the companies
that I speak with are looking for. All of the new hardware that they want to buy, they want to look at the open compute hardware because they're trying to run new applications and all the new applications are the things that you guys are developing. And they wanna run it on the same type of hardware that the hyperscalers are running it on. The same type of hardware that Facebook's running, that Alibaba's running and that Microsoft's running.
So ING, TCO, booking, 40% less energy than traditional Blade servers. So these numbers are important as you guys go get funding to go work on your other projects. The numbers are even more apparent for this Russian gaming company when you're looking at networking side.
The move from a traditional Cisco to a white box product is pretty significant. And it really matters when you're talking to the business side of the house about how you can justify the open source projects and moving to open hardware in addition to open software. The combination is really, really compelling.
So think about if I'm doing new applications, can I use open hardware to even increase my benefits? Standardization. Facebook deploys hundreds of racks a year. Thousands of racks across multiple data centers.
What's interesting about the way they deploy it is that if I'm gonna be doing a new application within Facebook or a new service, I have to fit it, I have to make sure that my application and my service runs on one of these seven types of racks. So every one of their racks, they just have seven types and they run at scale. Now if you go to a traditional bank,
they may have 3,000 different applications and 3,000 different types of racks. So always be thinking about standardization and this is the way that these guys can scale. When I talk to companies, they're always trying to get down to fewer applications and fewer rack types to increase that standardization.
I'm gonna touch on four of our embedded software projects. ONI, which is the Open Network Install Environment, it's run on probably almost every open networking switch. Switch abstraction interface, the Open Network Linux, and Sonic, the software for open networking cloud was donated two years ago by Microsoft.
It is the largest group that we have now in the open networking side. Most of your large companies, Tencent, Alibaba, Microsoft, all running Sonic. It's all available on GitHub and I encourage you guys to take a look at some of the things that are happening on the embedded firmware side.
The open systems firmware side is another project. There's a session tomorrow at one o'clock which Facebook is going to be talking about their Linux boot. I would highly encourage that. We are partnering very closely with the Linux boot side of things. Gundrala and Ron are both from Microsoft
and Google running the open systems firmware project within OCP. We also work very tightly with other open organizations. This is an example of the OPNFV virtual central office demo. So on the telco side, as people are rolling out 5G's, all the typical telephone exchanges
are being replaced by virtual central offices. I can run everything from a central office on software. I can no longer need high cost proprietary hardware sitting at my tiny little office or my tiny little demo rack out in the middle of nowhere. I can run that on all OCP gear.
And this is a typical encounter that we've had with the OPNFV organization. One last thing I always try to hit on is the efficiency side and why it's important specifically here in Europe. This is some legislation that was passed just a couple months ago around eco designs on servers.
There was a lot of issues happening, a lot of people against this, primarily large OEMs. This initiative had to deal with the idle energy usage of servers. So they focused on the idle piece of it because many, many servers in an environment are running at idle piece.
So in 2020 or 2025, 2023, there's gonna be new legislation around how efficient a server needs to be at idle. OCP and open designs are meeting this expectation. So here's an example of what OCP runs at idle. 52% more efficient than your standard OEM product.
So I think the takeaway is be thinking about, again, how can I use open hardware in addition to my open software efforts? If you wanna take a look, there's a marketplace. You can see all of the different types of OCP products that we have in our marketplace.
And more importantly, how do I get access to this? We have a company based in France. They have multiple partners around here. There's decommissioned OCP hyperscale gear. So instead of going out and buying brand new gear that may be costing quite a bit, all of the hyperscalers decommissioned gear is available through a partner network here.
This partner here, Circle B, is based in Netherlands. They can hook you up. You can take a picture of that in Meno or drop me an email at steveatopencompute.org and I'll be able to connect you. Or go to Nexity in Rapid Space. They have instance-based OCP gear, which is live right now.
And that's it. We have a summit coming up in October at the RAI in Amsterdam, where you'll see everything that you've ever wanted to see around OCP. And that's it. 25 seconds to go. Thanks again for the attention.
Any questions?