Re: [perf-discuss] [zfs-discuss] What causes slow performance under load?

2009-04-19 Thread Marion Hakanson
mi...@cc.umanitoba.ca said: > What would I look for with mpstat? Look for a CPU (thread) that might be 100% utilized; Also look to see if that CPU (or CPU's) has a larger number in the "ithr" column than all other CPU's. The idea here is that you aren't getting much out of the T2000 if only one

Re: [perf-discuss] ZFS performance issue - READ is slow as hell...

2009-03-31 Thread Marion Hakanson
james.ma...@sun.com said: > I'm not yet sure what's broken here, but there's something pathologically > wrong with the IO rates to the device during the ZFS tests. In both cases, > the wait queue is getting backed up, with horrific wait queue latency > numbers. On the read side, I don't understand

Re: [perf-discuss] why just one cpu gets loaded so heavily with 4 x 1GigE card?

2008-10-27 Thread Marion Hakanson
[EMAIL PROTECTED] said: > I am using sol 10 u4! > as per ur suggestion I tried following settings. > > set ddi_msix_alloc_limit=0x8 > set ip:ip_soft_rings_cnt=0x8 > > but seen processor servicing all the interrupts. Well, I certainly don't know why. I double-checked kernel revisions here, and f

Re: [perf-discuss] why just one cpu gets loaded so heavily with 4 x 1GigE card?

2008-10-24 Thread Marion Hakanson
[EMAIL PROTECTED] said: > see in your case too, cpu 12 is getting loaded, it will be interesting to see > what happens to cpu#12 when you increase the load further and increase > number of NICs. > > you can also use intrstat to get more details of what NIC interrupting what > cpu! once you reac

Re: [perf-discuss] why just one cpu gets loaded so heavily with 4 x 1GigE card?

2008-10-24 Thread Marion Hakanson
[EMAIL PROTECTED] said: > and how much you are seeing with 2 NICs? In some tests I ran today, it seems to max out at 150MB/sec, although it might have done better if the machine wasn't also duplicating an LTO3 tape over fiberchannel at the same time (:-). > did you also check with "mpstat 1 100

Re: [perf-discuss] why just one cpu gets loaded so heavily with 4 x 1GigE card?

2008-10-23 Thread Marion Hakanson
[EMAIL PROTECTED] said: > 1) Are you using jumbo frames in your setup? No, just the standard 1500 MTU. > With 1500 mtu what kind of throughput you see when all 4 NICs are used? We're only using two NIC's so far. We see 90-95 MBytes/sec on a single link with the current setup. Without the squ

Re: [perf-discuss] why just one cpu gets loaded so heavily with 4 x 1GigE card?

2008-10-23 Thread Marion Hakanson
[EMAIL PROTECTED] said: > 1) So, my question is why is cpu% so high in cpu#6 only? Though the problem > is solved with jumbo frames for 2 machines, if we increase number of nodes > this scalability problem will be seen with 3,4,5 ...machines (since cpu > utilization is very high with current state)

[perf-discuss] two filebench questions

2007-10-30 Thread Marion Hakanson
Greetings, (1) Where might one find the binaries (Solaris pkgadd, if possible) for filebench-1.0.0? Like several folks have mentioned, I've found the (incomplete) sources at sourceforge, and binaries for older versions are there. Have I missed something obvious? (B) (:-) Is there a