201–210 of 994 posts
They use Tofino as the networking switch. But Intel is discontinuing Tofino so I'm wondering what does future upgrade path look like for Oxide? Would they consider something like a p4 programmable DPU as the replacement?
Earlier quoted context omitted.
They mentioned it really quickly in their Oxide an Friends podcast but, IIRC, prices start at $500k. Some of the audience asked if they were going to do a smaller configuration like half or quarter rack. And they said they were looking into it but weren't sure the of the business case.
So the real question is whether 1 Oxide rack can outcompete 2 or 3 racks of normal commodity hardware. Or provide enough white glove after-sales support and written guarantees to peel away low end mainframe customers at a fraction of the price.
Given their management plane/API:
* https://docs.oxide.computer/api/guides/responses
the performance may be about the same, and CapEx as well (or maybe a little higher), but OpEx could be where you make it up in large(r)-scale operations.
And space efficiency is also not to be sneezed at: for some operations DCs/compute can be place anywhere because latency is that big of a deal, but in other places you need to be close to certain things (trading), and real estate can get expensive.
Earlier quoted context omitted.
No, there's just a lot more that can be said than they wanted to fit on this one page. Specs of the switch are available here: https://oxide.computer/product/specifications
I was asking about their NIC, not switch.
Earlier quoted context omitted.
The CPUs are x64 but the architecture is not that of a PC, there is no BIOS, etc. You couldn't boot Windows or Linux on the bare metal. The hardware, firmware and hypervisor are custom built for control, safety and observability. On top of that, the application OS all run on VMs which _do_ have a (virtual) PC architecture.
To me as a casual outside observer, the fact that they're using hardware virtualization at the top of the stack, after bcantrill gave so many talks about running containers on bare metal, is the most disappointing part. They could have had unbroken control and observability from the bottom of the stack all the way to the top. They got so close!
Can somebody explain in a sentence or two what this actually is and what the benefits are?
Weird there’s no mention of GPU at all when you’d think that’s what would prick up the ears of the hosting companies ….. stack a pile of GPUs in a rack, surely you can sell time on that. The Oxide machines seem to be aimed at 2020.
Nvidia GPU drivers are very proprietary, which means that admins and developers have limited visibility into them if they misbehave in any way. This goes against Oxide's philosophy of full visibility into a system that you purchase.
Nvidia's CUDA software has a significant lead ahead of AMD and Intel GPUs, and they're not going to open source it any time soon. But this is a rapidly changing landscape, and AMD and Intel and others are pouring an enormous amount of research into getting their hardware and software to match what Nvidia has going. Nvidia is in pole position, but they're not guaranteed to stay there.
There's still a large market for the CPU workloads that Oxide is offering. For now, Oxide will be concentrating on meeting this traditional compute demand. But you're right to point out that in 2023, the absence of a top tier GPU in these racks is noticeable. I suspect Oxide will want to include some form of GPU or TPU into the next version of their system, but they won't just grab whatever hardware happens to be in fashion. It needs to work with their system as a whole.
Earlier quoted context omitted.
If I have server under maintenance with Dell or HP, they would replace the server or component for me in such a case. Which would probably be a lot faster than trying to find someone who could maintain some non-standard firmware (as good as it might be). Even if I had to replace the server on my own cost it would probably still be cheaper. And it would be easy to replace because it's commodity hardware, that was kind…
Replacing hardware? Sure, they'll help. What about debugging firmware though? I'm curious how much help you would get from Dell fixing and patching complicated firmware errors. A side benefit of the openness is that firmware issues can be discussed publicly, and the patches can be upstreamed into the main repo and made available to every customer (and even competitor). This gives you the kind of network effects that…
If they replace the broken component with a working component, then I don’t care how they fix their firmware errors.
Earlier quoted context omitted.
Is that why you went to Berlin? How does Berlin compare to SV for cost of living?
This site is usually helpful: https://www.numbeo.com/cost-of-living/compare_cities.jsp?cou...
This is actually a pretty big deal. They sell servers, but as a finished product. Not as a cobbled together mess of third party stuff where the vendor keeps shrugging if there is an integration problem. They integrated it. It comes with all the features they expect you to want if you wanted to build your own cloud. Also, they wrote the software. And it's all open source. So no "sorry but the third party vendor droppe…
Proxmox and a full rack of Supermicro gear would not be as sophisticated, but end result is pretty much the same, with I imagine far far better bang for buck.
I like it, but it doesn't seem like a big deal or revolutionary in any way.