On Fri, Jan 26, 2024 at 4:45 PM John Stoffel <john@stoffel.org> wrote:
"Gregory" == Gregory Boyce via WLUG <wlug@lists.wlug.org> writes:
On Thu, Jan 25, 2024, 5:03 PM John Stoffel <john@stoffel.org> wrote:
> "Gregory" == Gregory Boyce via WLUG <wlug@lists.wlug.org> writes:
Coincidentally, I just recently started in a support role at Scale Computing, which provides virtualization leveraging qemu/kvm, with their own proprietary clustered filesystem.
Can we convince you to come and give us a talk? This looks really interesting!
I'm open to giving a talk, but I don't tend to be free on Thursday evenings.
I think we could schedule something if it works better for you. Is there a particular night that works better? Ideally I'd like to have you do it in person, if only so I can pick your brain better.
I'm typically free on Mondays and Wednesday evenings, as well as every other Friday.
And does Scale offer support for datastores (not sure the term) on NFS storage? Since that would make our transistion simpler.
No. We have a proprietary file system that distributes blocks across the cluster. In addition to redundancy of blocks across servers and disks, it also offers tiered storage where the bulk of content will exist on SAS HDDs, with more commonly used blocks moving to SSD or NVME storage. There's some good information here in the linked PDF: https://www.scalecomputing.com/resources/sc-hypercore-scribe-theory-of-opera... The VMs themselves would still be free to make use of any networked filesystems. -- Greg