I'm confused what their product is... "The cloud you own"? Isn't that just... a server? Sure, it looks like a very nice server, but is there anything special about it apart from that?
I'm confused what their product is... "The cloud you own"? Isn't that just... a server? Sure, it looks like a very nice server, but is there anything special about it apart from that?
Nothing particularly special. Their proprietary technology gives you some minor improvements in performance, reliability, and power efficiency relative to what you could assemble into a rack yourself. But more importantly for large enterprise customers they give you a single throat to choke: if something doesn't work then you can call them up to fix it with some assurance that it will get handled quickly. They won't point fingers at another vendor.
I listened to this recently which did a great job explaining the challenges that companies face when going 'on prem' and the hard problems that oxide is solving
https://newsletter.pragmaticengineer.com/p/the-history-of-se...
While I haven't looking into it all that deeply, I'd say it's a replacement for vSphere and cobbled together hardware and networking, all with a centralized management interface/API.
Traditional hosting still, to some extend, struggle to provide the API, on demand, drive requirements for modern developers, who expect to be able stand up a bunch of virtual machines in a minute or so, especially if you also want a new private network, maybe some IPs and storage pools.
Having a single provider for your entire stack, software, hardware and network avoids the annoying back and forth with vendors, blaming each other. Having just one support contract for your entire stack is a pretty large plus.
> Traditional hosting still, to some extend, struggle to provide the API, on demand, drive requirements for modern developers, who expect to be able stand up a bunch of virtual machines in a minute or so, especially if you also want a new private network, maybe some IPs and storage pools.
If you don't like vSphere (who does?) you can do all that in Proxmox.
Proxmox isn't quite there yet for scalibility and hyperconverged but it is getting there really fast. It's more of a competitor to Microsoft HyperV HCI.
Agreed. Its more reminiscent of Cloudstack or Openstack from what I gather. I'm thinking of Jetstream2, but for you buy it rather than rent some of it with an NSF budget.
Proxmox is nice and getting there but what you simply can't replicate is the lower level firmware and managing layer. Proxmox on commodity hardware will always be limited by that. Same goes for deep integration with the switch that Oxide can do.
It's more than a server, it's the whole rack with networking and all that, integrated and with unified management.
There is some company who for reason X and Y rather (or are obligated to) do on-prem for their hosting needs. But setting up a full (or several) racks, with all the required equipment for proper networking, storage, etc, can be quite the hassle. And if you want cloud-like functionality (completely API manageable virtual network, VM, storage pools, ...) it's another can of worm. Having a "plug'n'play" cloud-like system on-prem that do not require several engineers who know 10's of different vendors tech is definitely worth the premium for those company.
I think its the end-to-end, integrated nature of it.
API driven, have "elastic resources", etc, etc. Rather than bolting together various solutions you get to have a "Cloud-like" stack in your own datacenter.
Yep, it’s basically slick hyperconverged infrastructure that you can buy a rack at a time without a subscription license.
Do you get updates to the proprietary software stack if you go without a subscription license?
If the answer is no, then you might own the hardware on paper, but you don't control any of the software that makes said hardware useful.
If the answer is yes, on the other hand, then one must ask who is paying for those updates, because that can't be sustainable.
So while most of the software is open source rather than proprietary, you still have a fair point that customers pay for support (as they do with most enterprise products). One could theoretically use the product without first-party software updates, managing the open source oneself... but that would have practical impediments (and runs counter to the all-in-one simplicity that customers value in the Oxide product).
Two points about your last point. First, software improvements benefit all customers; as the business grows, the effective cost per customer shrinks. Also, most customers grow their Oxide deployment or will replace hardware after a depreciation cycle. The sustainability of investments into the software (and the product generally) is on solid ground.
Back in the 90s and 00s, lots of companies churned out software products that were sold once, supported forever. It was a sort of Ponzi scheme, supporting old customers with money from new customers. Which was okay during a period of high growth. But sooner or later the market matures, growth plateaues, and the cost of ongoing maintenance becomes a much bigger problem.
Right now you're growing fast and swimming in VC money, so this is probably not an issue. At some point, though, you might find that even hardware depreciation cycles don't provide as much of a cushion as you hope they will. In an economic downturn, people might suddenly realize that Oxide hardware actually remains serviceable much longer than they expected. :)
> Do you get updates to the proprietary software stack if you go without a subscription license?
what proprietary software stack? they just publish it all on https://github.com/oxidecomputer/ .
The software is open source and developed in the open. You can pay for support, but there’s no software licensing cost.
A stack of 38x 1U and a switch does not a cloud make.
Slightly less pithy: they're selling rack-scale systems, with power, hardware, network, and control plane software all integrated. Something that presents to the user as something more like API to interact with than a pile of servers to be managed.
Think "AWS in a box", with the high level features you’d get there. API, Terraform, "managed" products like load balancers, database, etc. You get all that in a turnkey rack delivered to the datacenter or office doorstep.
It’s a step higher than a vSphere/Proxmox cluster.
If you have ever struggled with a server whose bios won't netboot because there's a misconfiguration on the switch, or the Ethernet cable is not coded right for the speed of the server's card (because your vendor silently "upgraded" you to 25 Gbit because they were out of 10 Gbit cards), and then when it does boot, it is thermally throttled because it's tiny fans happen to be blowing in the one spot where your electrician tied a bundle of electric cables 10 cm thick, and then once you get the thermal throttling problem solved, you find out your version of IPMItool is incompatible with some stupid extension your server vendor defaulted to "on", then you might understand why Oxide is a good deal.
If you idea of installing a server is "terraform", you're not going to get it.
All that is fine and well, and I love coreboot, openbmc, etc. as much as the next guy, but how is this a business with growth or scale? In particular, you are not going to sell to the large clouds as they do a similar thing in house, you are not going to sell to the large LLM labs as there isn't much of a story with NVIDIA here. All you are selling to is on-prem deployments for old(er) school workloads, which to me is a shrinking market to begin with. You are like a fancier version of Dell or Supermicro. I don't get it. But maybe this is the Dropbox comment.
There are plenty of old-school companies in Europe still working on moving to the cloud. Now that there is a burgeoning movement towards avoiding American cloud providers, Oxide could have an opportunity to sell "private cloud" servers instead. If they play their cards right, they could make significant inroads in European markets.
I worked at Red Hat a few years ago... early 2020s. For our customers, something like 80% of our RHEL customers were still on-prem.
Yes, cloud is huge, etc. But there's a very big iceberg of on-prem.
Because cloud customers don't pay for Linux and they use Ubuntu.
Because when customers go from on-prem to cloud they throw away their distro tooling and support needs? I get it, I've worked at plenty of small and large cloud-native places, but saying "All you are selling to is on-prem deployments for old(er) school workloads, which to me is a shrinking market to begin with." you're not at all accurate. Or if it is shrinking, it is still a huge market, likely a much larger market than you think.
I think you might be underestimating how big the "old(er) school workloads" market is. And, at least from Oxide's point of view, it isn't shrinking but instead growing. A certain segment of tech has been enamored with the public cloud for the last ~15yr but personally over my career spanning that time I've seen some real drawbacks. "Spaghetti infrastructure" is a real, bad problem. Cloud pricing models heavily penalize some totally legitimate workloads. Keeping costs down while scaling up is really, really hard. If you own a fixed amount of hardware and buying more of it is expensive, you tend to use it more intelligently. Or maybe the one on-prem company I've worked at was just exceptionally good at computers?
> you are not going to sell to the large LLM labs as there isn't much of a story with NVIDIA here
Oxide just recently talked about that actually the LLM people do want to buy Oxide. Because turns out, doing everything around LLMs also requires compute, and quite a lot of it. And when you already have to deal with massive issues to run a complex advanced Nvidia stack you might not also want to worry about what firmware bugs Supermicro is delivering.
If you are not one of the hyperscaler who already has all the CPU based infrastructure on their own cloud stack (google, amazon, facebook) then Oxide is quite interesting.
Also as for this shrinking/small market claim. About 50% of IT spend is still outside of the cloud. While nobody know the real number, its still a gigantic market, much bigger then most people realize. And it might not be shrinking because the bad economics of cloud are becoming increasingly clear to many company. Along with other trends such as making computing more local, not letting US companies control everything.
> You are like a fancier version of Dell or Supermicro.
Dell has a market cap of 80 billion $, Supermicro has 20 billion $. Must really suck to be them I guess. I'm sure Michael Dell wishes he had done something worthwhile with his live instead. I mean he could have worked for Digital Equipment Cooperation instead then he might not have ended up being such a loser.
I feel you are being really dismissive talking as if aiming for that is somehow not worth doing.
>Dell has a market cap of 80 billion $, Supermicro has 20 billion $. Must really suck to be them I guess.
For a startup, if the thesis is to take market share away from those two, it's actually not such a good story. You need a product that is 10x better than the competition, and I'm not convinced that the enhancements to firmware, reliability etc. amount to a 10x jump in business value prop. You aren't making silicon. You are still ultimately a purveyor of other people's IP.
The claim that you need a 10x better product to win any market share is simply incorrect, both logically and historically.
Maybe if people that bought Dell had a deep love for Dell products and were deeply integrated unable to move, but even then 10x is a waste exaggeration.
But if you have any serious academic literature that underlies this 10x claim I'm happy to take a look.
> You aren't making silicon. You are still ultimately a purveyor of other people's IP.
And neither does Dell and they are worth 80 billion $. And AMD doesn't make semiconductors, so they relay on other people IP. And TSMC doesn't make their lithographic machines or many other things, relaying on other people's IP. And all those materials relay on other people IP to be brought to market in the first place.
This is just a silly argument that for some reason puts CPU design companies as 'the real deal' and everybody else is somehow not good enough.
Historically good systems companies make just as good margin as most CPU design companies, specially those that don't have near monopolies.
> it's actually not such a good story
They are making inroads in a market that is 100s of billion $ large and people invested 300M$+ in them because they see costumer demand. If that's not good enough for you then I don't know what to tell you. I wish any of the starups I have worked at that kind of opportunity.
It seems to me you operate in a sense where anybody that doesn't go for a monopoly in a 5 trillion $ market is somehow not 'worthy' of being a startup. That just a very strange perspective on reality.
Well, the monster wearing broadcom skin fucking over vmware licensees makes for a very interested market.
There's a lot of stuff that even if you put majority in the cloud, you want local deployment for security (inc. "operate when internet is out" security/reliability) and latency reasons.
For various reasons, vmware was pretty strong contender in this. Oxide racks are comparable in "sanity of mind" in deployments, and last time I was in a company that could use that the only major breaker was lack of ability to ship a raw VLAN to a VM, to enable direct replacement of existing vmware stack. But if it's not already fixed, it is not particularly hard to fix.
Surely vmware licenses are more easily replaced via Proxmox? Why would you care about Oxide, which is a hardware vendor?
Proxmox is fine if your vmware deployment was quite small. Single oxide rack at max density is going to similar values as official scaling sizes for proxmox, and very much isn't limit of what we did with vsphere.
And Oxide sells a complete hardware + software solution, including virtualization and SDN - essentially it's a physical equivalent of up to 32 node virtualization cluster per rack, with builtin SDN and SD-SAN, that already has features to combine for more.
Is this about limits to what the Proxmox folks will officially provide support services for, or perhaps what can be comfortably managed in the Web UI? These are very valid concerns either way, but the Proxmox VE software itself does provide a comprehensive API and that should scale quite a bit higher.
You're not wrong. This is old school enterprise dressed up as a start up.
Brian is trying to recreate Sun and using investor money to do so.
Good luck to them but I can't see it ending well.
… Ancient jvm running under wine with webstart just to get to the remote console? ;) I do not miss those days.
You mean you don't have that issue anymore? I'm jealous.
at least relatively – gen9s, almost 10 years old by now – modern HPE iLOs have HTML5 remote console. dell too I think.
webstart at least works if non-plugin. Flash not so much -- looking at vcflex.