4 ms·

What I have in mind is a platform to run virtual machines at home. This is definitely a niche product, but there's a significant overlap between nerds running k

What I have in mind is a platform to run virtual machines at home. This is definitely a niche product, but there's a significant overlap between nerds running kubernetes at home and people who have influence over what server products to buy for their datacenters. The form factor would be towers or NUCs, or something you could put under your desk without it sounding like a vacuum cleaner. The CPUs in Oxide racks are AMD, so, presumably AMD-based compute rather than ARM. The idea is to have a system that's as analogous to an Oxide rack as reasonably can be, but on a smaller, cheaper scale. Yes, Hubris is open source, but much of Oxide's value proposition is hardware integration. So the hardware would be at least vetted, if not completely designed from the ground up by Oxide and Framework. To be clear, I'm not making the argument that this product would be a slam dunk success for Oxide (or Framework). The benefits that I'm purporting might not be worth the cost of development. In the history of computing, a lot of the infrastructure development has tended to be bottom up rather than top down. Server racks are the descendants of PC's, not mainframes. This could be a way for Oxide to have something that aligns in that direction.

denotational 2y ago
Once you get away from the “rack” as the unit and go back to a tower/NUC, I don’t think Oxide has anything special to add, but I may be wrong since I haven’t used their product. A Dell (using them as an example since I’m familiar with PowerEdge) server/tower/NUC is pretty well integrated; the Oxide value add, as I understand it, is that your entire rack is integrated as opposed to being your own assembly of PDUs/servers/switches/etc. Perhaps their management software (for spinning up VMs, managing storage, etc.) is high value, but I’m not sure how that works when you lose the rack-level integration and go back to a tower/NUC. What do you see Oxide as adding at the tower/NUC level that takes you beyond the integration of a traditional tower/NUC from a vendor like Dell? I have seen similar excitement for “hobbyist Oxide” on other threads, but I’m still not completely sure what this actually means and how it would distinguish itself from existing sub-rack commodity hardware. A sibling comment mentions smaller racks, which is certainly something I could get behind, but that’s not the same as towers/NUCs (or maybe that’s exactly what you mean: something that looks like a tower/NUC but is really a mini-rack with multiple miniaturised computers/switches/etc. inside?). > The CPUs in Oxide racks are AMD, so, presumably AMD-based compute rather than ARM. These don’t run Hubris though; the x86 cores are running Helios [0], Oxide’s flavour of Illumos (i.e. Solaris) that includes a bunch of binary blobs that aren’t even available, let alone open-source: they plan to make these “public” at some point, but it’s not clear whether this means source-available. Based on the chips directory in the repo [1], they’re targeting a mix of NXP and ST parts, which are Arm, and the user isn’t likely to see them or care what firmware they’re running: I don’t think replacing the firmware with something running on top of Hubris is a value add, or even particular interesting for the hacker/hobbyist at all, in the sense that they’re a tiny cog in a much more complex system, and if you’re interested in hacking on something like that then it’s much easier/cheaper to do it in isolation rather than trying to do in-circuit debug on the part that’s keeping your server from catching fire (but that’s just my opinion!). [0] : https://github.com/oxidecomputer/helios https://github.com/oxidecomputer/helios [1] : https://github.com/oxidecomputer/hubris/tree/020d014880382d872d048fbfe1e8152a39e7c47a/chips https://github.com/oxidecomputer/hubris/tree/020d014880382d8...

denotational 2y ago

Once you get away from the “rack” as the unit and go back to a tower/NUC, I don’t think Oxide has anything special to add, but I may be wrong since I haven’t used their product. A Dell (using them as an example since I’m familiar with PowerEdge) server/tower/NUC is pretty well integrated; the Oxide value add, as I understand it, is that your entire rack is integrated as opposed to being your own assembly of PDUs/servers/switches/etc. Perhaps their management software (for spinning up VMs, managing storage, etc.) is high value, but I’m not sure how that works when you lose the rack-level integration and go back to a tower/NUC. What do you see Oxide as adding at the tower/NUC level that takes you beyond the integration of a traditional tower/NUC from a vendor like Dell? I have seen similar excitement for “hobbyist Oxide” on other threads, but I’m still not completely sure what this actually means and how it would distinguish itself from existing sub-rack commodity hardware. A sibling comment mentions smaller racks, which is certainly something I could get behind, but that’s not the same as towers/NUCs (or maybe that’s exactly what you mean: something that looks like a tower/NUC but is really a mini-rack with multiple miniaturised computers/switches/etc. inside?). > The CPUs in Oxide racks are AMD, so, presumably AMD-based compute rather than ARM. These don’t run Hubris though; the x86 cores are running Helios [0], Oxide’s flavour of Illumos (i.e. Solaris) that includes a bunch of binary blobs that aren’t even available, let alone open-source: they plan to make these “public” at some point, but it’s not clear whether this means source-available. Based on the chips directory in the repo [1], they’re targeting a mix of NXP and ST parts, which are Arm, and the user isn’t likely to see them or care what firmware they’re running: I don’t think replacing the firmware with something running on top of Hubris is a value add, or even particular interesting for the hacker/hobbyist at all, in the sense that they’re a tiny cog in a much more complex system, and if you’re interested in hacking on something like that then it’s much easier/cheaper to do it in isolation rather than trying to do in-circuit debug on the part that’s keeping your server from catching fire (but that’s just my opinion!). [0] : https://github.com/oxidecomputer/helios https://github.com/oxidecomputer/helios [1] : https://github.com/oxidecomputer/hubris/tree/020d014880382d872d048fbfe1e8152a39e7c47a/chips https://github.com/oxidecomputer/hubris/tree/020d014880382d8...