David Murphy wrote:
> > > > but in RAID5 you write different data to each disk...
> > > You have an extra column's worth of data to write - the parity
> > > info.
> > for which you have an extra disk. so forget that argument.
> No, you forget:
> 1) If the disks are all on one SCSI bus, you have to issue the
> requests sequentially - and wait for 3 writes to be acknowledged,
> instead of 2.
I haven't benchmarked anything, but it seems to me that RAID5 ought to
be at least as fast, if not faster, all things considered.
Whilst there are three writes, each of the three writes is only 50% of
the size of the two writes to two disks in a mirror. On the bus, the
three writes must be issued sequentially, but the drives operate on them
in parallel and your bottleneck ought to be the speed of seeking and
media writes. If SCSI or PCI bus bandwidth is your bottleneck, you're
probably in Happy Land anyway.
> 2) If you have three SCSI controllers, one disk on each, then you
> still have to copy the data to two controllers instead of three.
If you're worried about the data copy, you're copying 200% of the
original data for mirroring vs. 150% for RAID 5.
> 3) No matter what, you have to calculate the RAID5 parity info before
> you write it.
This is true, but it's most unusual for that to happen at a slower pace
than the disk or SCSI bus speed. You may lose a little in latency but
the bandwidth shouldn't be affected.
Paul.
Maintained by the ILUG website team. The aim of Linux.ie is to
support and help commercial and private users of Linux in Ireland. You can
display ILUG news in your own webpages, read backend
information to find out how. Networking services kindly provided by HEAnet, server kindly donated by
Dell. Linux is a trademark of Linus Torvalds,
used with permission. No penguins were harmed in the production or maintenance
of this highly praised website. Looking for the
Indian Linux Users' Group? Try here. If you've read all this and aren't a lawyer: you should be!