[OT] improve disk i/o on file server
Jurvis LaSalle
lasalle at bard.edu
Fri Jan 7 00:59:22 UTC 2005
The goal:
primary: Improve file I/O
secondary: improve redundancy if it doesn't sacrifice speed
The situation:
1 Dell PowerEdge 1550 with 3 Ultra160 SCSI bays running RHEL AS v3-U4
2 Seagate ST336706LC 36GB 10k drives
leaving 1 drive bay ripe for expansion
1 green sys_admin without a guru to turn to
Currently, I'm using LVM to pool the two SCSI drives in one volume
group. 47GB of the volume group is allocated and partitioned into
logical volumes, leaving 19 unallocated. None of the logical volumes
are striped. There is no redundancy- the server is backed up to tape
regularly.
I've budgeted to fill the third SCSI bay. My original intent was to
buy a similar third disk, pop it in, use vgextend to add it to my
volume group and then very carefully dump and restore the old logical
volumes to new ones striped across the three drives. Times have
changed however and I can afford to buy drives that are faster and much
bigger than the original two. And here I started thinking...
I could get a drive twice as large and RAID 0 the two smaller disks and
then RAID 1 the results together and put LVM down or... the scenarios
are endless! I realize that I'm just going to have to buy the drive
and test each scenario I can think of.
The questions:
Long post short (too late!), I'm looking for pointers (anecdotal tales
welcome) about mixing SCSI drives of different speeds (Ultra160 v.
Ultra320) and mixing different rotational speeds (10k v 15k). Finally,
any suggestions about the best setup are also welcome.
yaiga,
Jurvis LaSalle
PS I ask about mixing U320 with U160 because most of the 146GB SCSI
drives in my price range are U320/10K and google returns contradicting
info about putting U320 in a U160 bay. Thanks again.
More information about the redhat-list
mailing list