 So we hear the linear connect and who are you I'm Edville Medi from packet. I run the works-on-arm project for packet So you just had a presentation and you were talking about the latest stuff that's going on So there's a lot of different platforms that you have. Yeah, so Most of the system that we run for arm is Cavium Thunder X based And we have in addition some high silicon high 1616 based hardware the Do5 platform that Huawei built and in addition we're in the midst of Adding to that some Thunder X2 systems as well as a couple different Platforms from socio-next their sync waser. I think that's how you pronounce it and also The latest is the ampere Equipment that just got announced today So So the Thunder X2 for example you get them as soon as they became available kind of or they like they took you out to become available Or we got them pretty early The The challenge with new hardware is that you have to shake out the firmware So we got some systems. We just upgraded to the latest firmware that they have And are now in the midst of getting them into the hands of some of the Some of the projects that we work with that really need that hardware because it's special It's the only System on the market that does large systems extensions So we have I spoke with someone today who's very interested in testing out some compilers that were enabled for that So There's a lot of people waiting for this at Thunder X2, right? It's supposed to be a super powerful It's a very interesting machine. It has it can be configured with Let's see so 228 core CPUs With four hardware threads per core so a total of on the order of 215 or 216 or 224 threads and most People writing software for arm don't think in terms of systems like that, right? They're their worldview is four core machines eight core machines 16 core machines so it's it's very interesting for software developers to to Work to figure out how to take the best advantage of that sort of system so So it's called packet that net and it's not a very very old company, right? But it's very active and very Love things happening. Yeah, so we've been in business for about four years now packet just announced $25 million series B round Which I should know who the lead investor is, but I don't off the top of my head But it's it's been I've been at the works on our project for a little more than a year now and it started out really with Understanding that we knew that people needed to get access to hardware But we didn't quite we're quite sure who it was going to be and it has been a bunch of work to Understand the ecosystem make friends with people with distributions get referrals into language development and cloud native computing efforts and telco workloads and various things to just try to Really understand the infrastructure so that we can help at the best So you the world leader is packet the world leader and an arm server kind of making it available We're clearly the world leader in bare metal arm servers in terms of access to the public The packet is small compared to an Amazon or compared to Google or These huge Microsoft right these companies in somewhere in Texas. Yeah giant things just enormous, right? So we're not at that scale But we're innovative and pretty quick moving and have had a lot of luck working with people who for whatever reason their problem lends itself to working inside a physical bare metal machine Rather than trying to solve their problem inside a virtual machine. So The packet is kind of new leader and this kind of bare metal or not really leader. I would say we're a leader So there's only I Guess it I saw someone who did a market analysis of like where can you get bare metal? How much it cost per hour? How how how many Davis centers there are people in what not and We're definitely one of the larger companies in what's a relatively small market I would say that the need the appreciation of what you can do with Really understanding what the hardware does and taking the most instead of dealing with a virtualized view of the world Understand what the hardware can do and write to that hardware Can be really powerful for some people We've had experience with people moving workloads from virtual to bare metal and they you know Get better performance and it costs less Without without too much extra work. So that's that's been a positive experience So but then when we talk 86 and an arm so it's a successful company. That's why there's a service bees happening and all that stuff, right? So it's growing. It's growing. There's a lot of funds to expand where there's demand We're in the midst of opening up a new data center in Texas in Dallas And expanding out the size of the team. So it's it's been a good you know, it's it's heading in the right direction and The what's what's been the experience with the arm stuff so far You were mentioning that there's a lot of people that do interesting things you didn't maybe didn't expect So the experience with arm has been If I compare where we are now to where we were a year and a half ago a year and a half ago The distributions were in good shape and a lot of the user contributed software was was not in as good shape and what we've seen is a series of Investments by various folks to make sure that not only does the software That they're trying to use that it's portable enough that it works on arm but also that people have been optimizing their code so that it runs fast on arm and What I've been looking for in the works an arm project is a diversity of enough different projects going on right some of the most important things and also some more I Don't know esoteric obscure Not very well known projects to make sure that we get a really broad coverage of the whole ecosystem And you're mentioning that you have some things that are paid at market price and some things are free What is free? So the the works on our project is sponsored by arm To give access to open source projects to get infrastructure for Test and development and primarily for CI CD So the people who are getting that free level of access are doing things like producing The latest version of Java or the latest version of go or that are doing operating system development or some key application development Or they're running CI systems aimed at free service for folks But fairly quickly people are especially who are running CI Realize that there's a commercial market as well because they can do a system that will do Builds for embedded systems and other sorts of 32-bit chips Much like six to eight times faster native compared to doing it emulated So there's there's a lot of value in that and they're willing to They're willing to pay for that service. So you say for example, there's some companies that that build Build what's called the image for IoT devices, right? So there's a company called resin which has been doing that they're using packet as part of their infrastructure and Using resin to manage essentially you're managing containers on a embedded Linux system through the resin infrastructure the whole build test Deploy manage maintain cycle. So I've been doing our server videos for eight years. No, yes So is it just about to explode? What do you think? You know the the biggest challenge that we had at packet is availability of hardware There's a lot of stuff that's been announced that's very hard to buy and we still run into genuine issues sometimes of taking servers from vendors and Integrating them into our infrastructure I have long since given up on predicting when markets are going to explode I know that there's steady demand. There's more demand for the service That I provide for people doing build and development work Every every month there's people coming people who I don't know Coming to me and saying hey, we've got code. We want it to work on arm We understand that you have some infrastructure. We'd like to support arm in our product That's all that's all in a really good direction now. Are are they supporting that for? Embedded systems are they supporting it for single board computers? Are they supporting it for servers? Are they supporting it just because they think it would be good to have a portable product? hard to say so So you would not you don't want to say on camera if you think that next your arm will be over by 2020 There's gonna be 25% of all new servers That was two or three years ago. That's what I'm said. Yeah, so I Think it is up to a bunch of server vendors to really execute on that demand We know that there is Good uptake in the HPC market for arm-based systems for a variety of reasons We know that arm has a very strong position and embedded It's but cracking the server market is a really I don't want to minimize how hard that problem is So I'm not gonna guess. I'm very hopeful, right? I really love to see a 25% market That'd be great because it's great Not just because you've been doing all that stuff that has arm related in the name, right? But it'd be great because it's it's exciting for the the whole industry Competition happening in something different, right? Yeah, it's it's always exciting when people bring competitive products to market that are That address this well understood needs, but they're different enough that people can you know, I have to make a decision about stuff I think it's good for I think it's good for the arm world to have strong servers in that world even if the demand for them Doesn't hit 25% of the market. It's still good for the ecosystem that people can develop for arm-on-arm Right that they can not have to cross host their platforms for for development and you have some Qualcomm servers and you have some But not everybody can just access it yet or something it sounds like because It's been nearly a year that I kind of heard them say. Oh, yeah, it's available The Thunder X2 and the Qualcomm, but it's right. It takes completely available. It takes well. It takes time From the time that someone announces working first silicon to the time that you can go into mass production that If you decide to go into mass production it takes time you have to Know early silicon is often different from final silicon. There's changes that get made as people find issues Um There can be In a lot of these systems that have a lot of cores you have a lot of choices about what Exact part to make do you make the? You know 48 core part you make the 46 core part you make the 44 core part. How much did these things cost? What are you yields on all the prong on these things? So there's from the time of announcement to the time of real Quantity general availability. There's a lot to be figured out And then some of the sales cycles are just longer, right? People are not selling these one at a time For the most part, I mean sometimes you can get you can order one But a lot of the server chip makers are really targeting the hyperscalers of this world where You're looking at you know data center size purchasing rather than One at a time purchasing So part of the the work that you do is also Kind of promotion or what do you call it? There's you have a Twitter. Yeah, so part of it is just keeping track of the whole ecosystem I produce a newsletter almost every week that has news about It's sometimes has videos from you. It sometimes has press releases from folks Software announcements just general news about the ecosystem and then I maintain a Twitter account that will repost or Comment on various things going on. They're both called works on our arms They're both called works on our works on arm And so what's kind of like this some of the latest couple months of cool news there? I heard a couple days ago at her and on something. Yeah, and Per announced that they're have a new 32 core arm chip running at more than 3 gigahertz Which will be which is the fastest clock speed that anyone has announced so far I haven't seen that hardware up close to know exactly how it behaves. I don't know custom design Custom design course says this is the x gene 3. This is the x gene 3. Yeah, or something Related to the x gene 3. So it's the x gene. They're calling it e-mag now Okay, whatever reason people pick names. It sounds of something like because the American company With a joke, but yeah, there's I think mag is something to do with guns, right? you know it has It could be magnetic I could be any any of a number of things right very common. So I I guess the the Biggest news is usually small news because I I have 600 words to write every week or a thousand words to write every week So try to keep people abreast for smaller things the the biggest software improvement that I see Of late is a recent release of go the go language Which has a bunch of arm specific? optimizations in it That will make some code 20 30 40 percent faster just because they're using assembly instead of instead of go to Do critical path for some algorithms. So who's that who's the biggest users of go? Google is a big user of go and then there's a whole mess of what gets called cloud native software So Docker Kubernetes a bunch of Kubernetes tooling is written and go It's a very That's a very modern systems language it it knows how to deal with systems that have More than one CPU fairly well Whereas some older languages struggle to do parallelism across a bunch of across a bunch of threads so it's in it's a It's not the only language that you want to write code in but for some fairly substantial set of problems that are Characteristic of modern computing. It's a good. It's a good choice. So that's important news because So is a lot of Google infrastructure and go or did they say yeah a lot of Google infrastructure and go in Kubernetes? Which is a derived from a Google project? Is all written and go so so that could potentially be like a kind of what you call it a clue as to maybe They're on the verge of announcing huge arm data centers You know again, I'm not going to guess what they're what they're doing. What I would say is that It's been great to work with the developer community We get contributions from arm directly from Lenaro from a number of independent developers from various licensees like Qualcomm and and Cavium and and and others and just to see how people can Cooperate to get things done. It's been a helpful