Search results

I'd like to see how high you can go with an EPYC 7002. Anyone plans to get one and test how high it can go?

Moved the SAN to our production cluster - some new results available in the wiki efschu/AP-HA-CIAB-ISER

Which gives you 57% render advantage while having 100% more raw power - IMHO not impressive ;-) But nice toy if you do more then one job parallel.

As EffrafaxOfWug said, it does not scale beyond 12 CPUs, so this is benchmarking 12 Broadwell-E Cores vs 12 Skylake cores - meanwhile 32 Broadwell cores and 16 Skylake cores are idle'ing

So with this knowledge this platinum actualy doesnt look too good compared to his predecessor.

Are you sure your software can handle 88 threads? You may run the benchmark again with only one cpu active? - also this would give a good indicator on how it actualy scales per thread.

In your UEFI you should have an option to select primary video card - on my Gigabyte X370 I can only set that option properly if I have a second GPU in my 16x Slot, but after I have selected to use the 1x it works flawlessly w/o a GPU installed on primary pcie slot.

Single runs with qd1, qd10, qd256 and a parallel run oft two VMs with qd256. If your data is important you MUST use sync=always, else you loose the data which is not written to disks in case of failure. Officialy Esxi and scst can bypass sync requests, but not every Software is handling critical...

use a VM as gateway and passthrough the card - not perfect solution but should give you enough speed anyway

Reading the changes and came to the qemu point, then I'm reading qemu4 changes and see this: Q35: advertised PCIe root port speeds will now optimally default to maximum link speed (16GT/s) and width (x32) provided by PCIe 4.0 for QEMU 4.0+ machine types; older machine types will retain 2.5GT/x1...

I think there exists a dual ported 100GB DC P4800X

If I compare these results with the ones here (with optanes) on page 51 http://napp-it.org/doc/downloads/optane_slog_pool_performane.pdf I'm quite fine with that. If I run iometer I can easily reach 100k+ iops, but the performance is limited by my old initiator testmachine. I will post some...

There are 10 (dual-ported) u.2 nvmes, which can be accessed from both nodes. I have played around with the raid setup and actually got better results on 3x3 striped raidz + 1 spare, but the performance of raidz2 of 10 disks is way enough for my usecase, so I ended up in going for the higher capacity

worklog and some resluts @the end efschu/AP-HA-CIAB-ISER

efschu/AP-HA-CIAB-ISER here my worklog and some bench results @the end

I know it's an old thread but the answer is yes you can. I'm doing so with ocf:heartbeat:IPaddr2 and RoCEv1 with Globalpause.

I would like to use scst (actually I would like to use LIO...), but havent found any good (working with ubuntu ootb w/o doing debugging all time long) ocf heartbeat script, so I'm left with tgt I guess... Maybe I will write one by my own, but well dunno when I have time to.

Mellanox suggests in using MTU size of 4200 for iser, as I'm doing. This setup is on ubuntu 18.04, tgt and scst is working fine, but esxi is reporting the exact same errors with OFED 4.3 as with 4.5 on LIO. So I skip LIO cus I'm out of ideas (I think there is something going on with roce v1 and...

I usualy use nmon for that. nmon -> M -> M -> M

Downgraded OFED from 4.5 to 4.3 and scst is working. Trying LIO tomorrow.