perf are related to bandwidth throughput available on one disk and on fiber channel : you are limited to around 200MB throughpout per disk but you can go up to around 2GB on fiber channel then, if you can manage (that's what netapp does) to pull data out of several disks on the fiber channel bus, you can increase the natural throughput of fiberchannel disks this is not an exclusive netapp feature, it is a hardware trick, nowdays made as a standard called the striping (also named raid 0) so in theory , you could seriously manage up to 10 (or little more) disks in a volume in order to get good striping performance remember that you get perf increase if you double the disk split : getting from 1 disk to 2 lets you get a real perf improvement from 2 to 4 disks, from 4 to 8 and so on as netapp use a parity disk (raid 4) one disk can't be used for perf improvment, that's why i guess netapp advice at least 3 disks in a volume
bye
Thorsten.Lemke@computacenter.com wrote:
In the advanced administration and troubleshooting guide netapp indeed points out that a raidgroup should not have less than three data disks. Otherwise you will have a decrease in performance especially with read/write of large files. It has to do with the wafl striping mechanism but unfortunately the explanation in this guide is quite marginal. Perhaps someone from netapp can give us a more detailed answer?
thorsten
arden@nortelnetworks.com@mathworks.com on 23.07.2003 19:02:07
Sent by: owner-toasters@mathworks.com
To: toasters@mathworks.com cc: Subject: Mirroring using raidsize=2
Our company (in its wisdom) has decided that all "critcal" data must be on mirrored filesystems, not on raid.
On filers, if you set the raidsize to 2, you effectively get a mirrored volume.
Does anyone do this? Does it cause a performance problem? Is there a limit on the number of raid groups in a volume?
- Bruce
-- Bruce Arden arden@nortelnetworks.com barden2@csc.com CSC, Nortel, London Rd, Harlow, England +44 1279 40 2877