Re: Cluster and attached RAID performance
- From: "Bill Bradley" <wdbradley3@xxxxxxxxxxx>
- Date: Sat, 1 Jul 2006 06:08:20 -0600
I will, thanks. Our customer, and I, are PO'd about this. Dell didn't
mention this...
"Geoff N. Hiten" <SQLCraftsman@xxxxxxxxx> wrote in message
news:eQW4ikLnGHA.4496@xxxxxxxxxxxxxxxxxxxxxxx
Look at the HP MSA500 series SCSI enclosures. They use dumb SCSI
controllers in the hosts and put the smarts and the cache inside the
enclosure. I wouldn't call it a full SAN, but it is way ahead of the
basic hot-swap external enclosures most vendors sell.
--
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
"Bill Bradley" <wdbradley3@xxxxxxxxxxx> wrote in message
news:ueSi8pAnGHA.1244@xxxxxxxxxxxxxxxxxxxxxxx
Soooo...what do you use, if you need more space than what you have in
your server, and also need redundancy? Do you HAVE to go Fibre, or,
what?
Thanks.
"Geoff N. Hiten" <SQLCraftsman@xxxxxxxxx> wrote in message
news:%23ytiY%236mGHA.1244@xxxxxxxxxxxxxxxxxxxxxxx
Yep. Write cache only affects write operations. Read cache can't be
corrupted during a failover so it isn't turned off.
--
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
"David Day" <DavidDay22@xxxxxxxxxxx> wrote in message
news:1151603866.221174.135690@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
I am setting up almost the exact same configuration. The only
difference is that the PV220 has 13 15K 146G drives, divided into
Quorum and Data drives, configured RAID 5 Read performance is ok, but
anything with a lot of writes is 50% slower than my old system at
least.
Geoff N. Hiten wrote:
I am surprised that you are getting even 50% of the stand-alone system
performance. A PV-220 has no internal cache, instead it relies on the
cache
on board the PERC4 cards in each host. Since the hosts are clustered,
all
write cache is turned off on the PERC cards to avoid data corruption
during
a failover. This is why I don't recommend SCSI based clusters except
when
you use enclosures with embedded cache.
--
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP
"Bill Bradley" <wdbradley3@xxxxxxxxxxx> wrote in message
news:uMPe8a3mGHA.4212@xxxxxxxxxxxxxxxxxxxxxxx
I have a SQL cluster consisting of the following:
2 ea Dell PE-2850 servers, dual 3.2 GHz CPU, 4 GB RAM, 2-36
GB/RAID-1
for C:, same for D:, all drives U320/15K. running Win2K3EntSP1 and
SQL2K5SP1. Connected to cluster with Dell PERC4/DC in each server
1 ea Dell PV-220 cabinet (up to 14 drives, 13 when clustered)
with
following discrete arrays, all U320/15K:
2-18 GB/RAID1 Quorum Volume
2-36 GB/RAID1 LOGS
2-36 GB/RAID1 MSDTC
6-73 GB/RAID5 SQL DB
1-73 GB HotSwap
Original server was a single Dell PE-2650, dual 2.8 GHz CPU, 4 GB
RAM,
5-18GB/RAID-5 U320/15K, running Win2KAS and SQL2KSP4.
Performance of the new setup is 50% of the old one! User is pissed.
I was told a few things:
Don't mix drive sizes (but, I thought, they are in discrete
arrays, NOT
JBOD)
Too many drives on one RAID controller (no choice on that, only
one
connector from each server).
Anyone have ideas on this?
Thanks!
.
- References:
- Re: Cluster and attached RAID performance
- From: Geoff N. Hiten
- Re: Cluster and attached RAID performance
- Prev by Date: Re: SQL 2005 SP1 install into an Active/Active cluster
- Next by Date: Re: Cluster and attached RAID performance
- Previous by thread: Re: Cluster and attached RAID performance
- Next by thread: Re: Cluster and attached RAID performance
- Index(es):
Relevant Pages
|
Loading
