I seem to recall reading in a book cerca Solaris 10 about the kernel internals that it supported UNSIGNED SMALL INT (about 65k) cores but I am having a hard time finding anything to back that up and I gave that book away in a donation pile many years ago so I can't double check easily.
I know my loose recollection isn't that helpful but it jives with the other folks suggesting that the limit is way higher than BSD :) On Fri, Jan 8, 2021 at 12:01 PM Chris <[email protected]> wrote: > On 2021-01-08 08:39, Andreas Wacknitz wrote: > > Am 08.01.21 um 17:15 schrieb Chris: > >> On 2021-01-08 01:17, Joshua M. Clulow via oi-dev wrote: > >>> On Thu, 7 Jan 2021 at 23:53, Chris <[email protected]> wrote: > >>>> I'm looking to help building packages. Both current, as > >>>> well as newer versions. I have a large server farm. But > >>>> primarily BSD based. As such I'm looking into a new build > >>>> box, based on an Opteron or Xeon. So before I take the > >>>> plunge. I was hoping to add as many CPU/cores as possible. > >>>> Which begs the question: haw many CPUs does OI support? > >>> > >>> I don't think you'll be able to buy a machine with more CPUs than we > >>> support in illumos. I've personally used older HP machines with 4 CPU > >>> sockets and thus a lot of cores, and modern AMD systems which have a > >>> still surprising number of cores in one or two packages. > >>> > >>> In the unlikely event that you hit a problem based on core count, I am > >>> sure it will just be a bug that can be fixed. > >> OK so 4 sockets @16 cores/ea, or 2 sockets @64 cores/ea won't be a > >> problem then. > >> Thanks for the reply! :-) > >> Time to go shopping. > >> > >> > >> --Chris > >>> > >>> > >>> Cheers. > >> > > Tell us when your server exceeds 384 Cores / 3072 Threads. AFAIK that's > > what actual Fujitsu/Oracle M12 provide at max. :D > Sweeeeeet! :-) > TBH the only reason I brought it up was that we had a similar question on > one of the FreeBSD lists. Where they had difficulty exceeding 96 on a 128. > I've forgotten whether it turned out to be config of (actual) limitation. > Mind you that was BSD. But given the shared heritage, it seemed an > appropriate question. > Thanks for the stats. Maybe I should give Fujitsu a look. :-) > > --Chris > > > > > > _______________________________________________ > > oi-dev mailing list > > [email protected] > > https://openindiana.org/mailman/listinfo/oi-dev > > -- > ~40yrs of UNIX and counting > > _______________________________________________ > oi-dev mailing list > [email protected] > https://openindiana.org/mailman/listinfo/oi-dev >
_______________________________________________ oi-dev mailing list [email protected] https://openindiana.org/mailman/listinfo/oi-dev
