We've got a wide variety of HP servers as clients, running CentOS 5 and OpenISCSI, talking to three LeftHand Networks NSM2120s (which are essentially rebranded HP DL320s servers with their custom OS loaded on them).
We thought it was a switching issue as well at first, where we had a pair of Force10 S50Ns that I absolutely hated (still do, for lots of reasons), but after replacing them with a pair of Cisco 3750Gs, I can say that for as much as I hate the F10s, they weren't the cause of this particular problem. D On Oct 30, 2008, at 4:18 PM, Joshua Nichols wrote: > What hardware are you using? A colleague of mine just spent a few > weeks working to get peak performance out of some Dell md3000Is to > keep up with the needs of a client of his. Let's call them "an online > media company." > > My point is that baseline performance is determined so much by the > specific hardware, that what works to get an md3k cruising may prove > absolutely useless in your setup. > > Also, the switch is probably a bigger factor in overall speed than one > might guess at first... > > > On Oct 30, 2008, at 3:07 PM, Derek J. Balling wrote: > >> I'm trying to determine what's a good "baseline performance level" >> for >> my iSCSI SAN. We're seeing what we consider to be suboptimal >> performance, but I'm not sure where to pin the blame down -- or even >> if there IS any blame, maybe our metrics are in line with what people >> are getting for Linux iSCSI clients. Who knows. >> >> What I'm seeing is that - if I use IOMeter as my performance testing >> tool (which seems to be fairly common in SAN testing), I can only get >> about 300-400 read IOP/s, with total throughput of about 10MB/s. >> Contrarily if I let IOmeter test the local disk, it's 6x faster (1900 >> IOP/s, 60MB/s) >> >> Now that SEEMS awful low to us. There's a number of factors that >> might >> be contributing to it (it could be a network issue, it could simply >> be >> that the SAN hardware itself is saturated for IOPS -- I tend towards >> that last, but unfortunately, the SAN hardware we're using has >> NO(!!!!) means of letting me determine what the current IOPS are on >> the SAN disks themselves.... crazy-talk, I tell you). >> >> Anyone out there want to share their metrics? >> >> Cheers, >> D >> >> _______________________________________________ >> Tech mailing list >> [email protected] >> http://lopsa.org/cgi-bin/mailman/listinfo/tech >> This list provided by the League of Professional System >> Administrators >> http://lopsa.org/ > > _______________________________________________ > Tech mailing list > [email protected] > http://lopsa.org/cgi-bin/mailman/listinfo/tech > This list provided by the League of Professional System Administrators > http://lopsa.org/ _______________________________________________ Tech mailing list [email protected] http://lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
