Hey folks,


Been looking at a largish app that has scale up and scale out bits. Based on
experiments run on a 4-core z13 and 20 core Power8, I'd guess we could run
this app (with a bit of headroom) with ~24 z13 cores. Lots of memory would
be good (1T+) and, for scale out we need to be able to push 300MB/sec
presumably through HiperSockets or Unix sockets depending on how we decide
to split up the app. None of this sounds scary to me (of course, I'm not
paying for the hardware) but just curious if many people have z/Linux
experiences at this scale or higher.



I know this is all rather vague, and I understand lies, damn lies, and
performance statistics, and all that, but it would still be useful to know
if people are running z/Linux apps with 24 or more cores and how that's
working. I'd assume z/Linux can handle it with aplomb but worth knowing if
there are any dark clouds on the horizon. Has anyone had to mess with CPU
affinity or anything esoteric like that to get decent performance out of a
many processor system? People using larger page sizes for lots of memory?



On a related note, what kind of I/O rates are people able to get from
z/Linux. Which filesystem is being used at these rates? I'd assume high I/O
rates => SSDs or at least very heavily cached disks?



Thanks



----

Alex Kodat
Senior Product Architect
Rocket Software






----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to