On Tue, 20 Aug 2002 11:54, Philipp Schmidt wrote: > within this discussion, i got the idea to put an external journal for > the ext3fs on an raid1-volume the real data on a raid5, hopefully, when > writing the journal out to the disk having more data to be written an > once - this would be worth a try but i don't have the hardware to try > this and bad expirences with software raid5.
Ideally if using data journalling then the transactions would be large enough to encompass entire stripes and remove most of the RAID-5 penalty for small writes. However having an external journal on a RAID-1 array that's used for nothing else would really help performance. Also make sure that you are using the fastest part of the disks used for the RAID-1 (use ZCAV to measure it). The start of the disks seems to always be the fastest for IDE disks, but apparently SCSI disks sometimes do things differently so a benchmark would be a good idea. For best performance investigate some sort of battery-backed RAM disk device. Another issue of RAID-5 is that when writing a stripe if the power fails it may write some blocks of the stripe but not all. Having journalled data and an external journal on a non-RAID-5 should solve that. -- I do not get viruses because I do not use MS software. If you use Outlook then please do not put my email address in your address-book so that WHEN you get a virus it won't use my address in the >From field.
