Date: Fri, 19 Mar 1999 14:22:00 +0000 (GMT) From: Nick Hilliard <nick@iol.ie> To: freebsd-scsi@freebsd.org Subject: dpt raid-5 performance Message-ID: <199903191422.OAA27845@beckett.earlsfort.iol.ie>
next in thread | raw e-mail | index | archive | help
this particular topic comes up fairly regularly, but: I have a DPT 2044W RAID card (this is the Smartcache IV system with a RAID add-on module). The card has 64M non-ECC onboard RAM and is attached to 4 x barracuda 9Gb disks in a eurologic disk shelf. The cable is not long (1.5m), and the disk shelf is terminated mid-way on the shelf backplane, so the total bus length is pretty short. The terminator is the one recommended by Eurologic. The card is in a P-200 with 128Mb RAM and 3.1-RELEASE. Softupdates are enabled. If I set up a raid-5 system on these disks with a slice-size of 512K, the write performance is terrible. Bonnie reports: -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 256 541 7.5 491 1.7 458 2.3 4293 59.1 4335 16.2 193.6 6.8 RAID creation and newfs take forever, but the disks never appear to be particularly busy. iostat suggests that the system is managing between 30 and 50 tps. Copying files on to the raid with tar takes forever: for long periods, the write rate would rarely exceed 250K/sec, according to iostat. In comparison, bonnie reports 2.1Mb/sec write speed when using the same system configured with RAID-0/512K. I'm aware that RAID-5 means piles of parity/verify overhead, but this sort of speed hit seems ridiculous and looks as if there's a serious problem somewhere. Does anyone have any ideas? Nick Hilliard Ireland On-Line System Operations To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903191422.OAA27845>