Re: disk latencie problems on a brand new exchange server with 2 mailboxes!!!!!!!!
- From: "John Fullbright" <fjohn@donotspamnetappdotcom>
- Date: Tue, 23 Oct 2007 10:15:32 -0700
Figure a mirror for the OS, a mirror for the DBs, and a single drive for the
logs (or a single drive for the OS, a mirror for the logs, and a mirror for
the DBs depending on what fault tolerance trandeoffs you want to make) you
might be able to support 100 users at 1 IOP/user. There's no way this
system will support 1000 users.
From the description of the problem, odds are all 5 drives are in 1 RAID 5array with the OS, logs, and DBs in seperate logical drives (or on the same
drive; from a performance perspective it's all the same spindles). If this
is the case, I would expect you to be able to support more than a handful of
users. In fact, is sounds like 3 is too many. When you size disk, space
isn't the only factor; you have to consider performance requirements as
well. RAID 5 is a bad choice for Exchange 2003. Although it maximizes
space, performace is less than half that of mirroring (RAID 1 or 10). In
particular, write performance is grossly inadequate for and application with
a 3:1 or 2:1 read/write ratio. RAID 5 is appropriate for applications that:
1. Require a lot of space.
2. Write data infrequently
3. Read data mostly
These critera descibe some file servers. These criteria do not describe
Exchange. Back in the Exhcange 5.5 days, when you had an 8:1 read/write
ratio, you could squeak by with RAID 5 for the databases. You still put the
logs on a mirror. In a modern Exchange 2003 environment with Outlook clients
caching, you read/write ratio is more like 2:1. The RAID 5 write penalty is
way too high.
John
John
"Mark Arnold [MVP]" <mark@xxxxxxxx> wrote in message
news:sh7sh3110fmui171q55cbmcn7bdmvb72n2@xxxxxxxxxx
On Tue, 23 Oct 2007 10:55:13 -0400, "Jef" <nospam@xxxxxxxx> wrote:
we have an old exchange 2003 server that recieves alot of disk latencey
problems using MOM 2005. We just purchased a new Poweredge 6850 16 gigs of
ram quad core 3.66 with 5 15k hardrives. I moved about 3 mailboxes to this
unit and i am recieving disk latency problems (see below) from the new
server now also. Why is this happening, will this get worst when i move
all
1000 mailboxes to it?
High Disk Read Latencies for the past 10 minutes
PhysicalDisk: Avg. Disk sec/Read: _Total value = 5.49850987944241E-02. The
average over last 10 samples is 0.0549851.
You haven't got anywhere near enough disks in my opinion. Sure,
they're fast enough but five disks? How have you got them carved up?
If they're in a single RAID5 partitioned into drives that wouldn't be
the best way.
.
- Follow-Ups:
- References:
- Prev by Date: Re: Backup SMTP
- Next by Date: Re: Backup SMTP
- Previous by thread: Re: disk latencie problems on a brand new exchange server with 2 mailboxes!!!!!!!!
- Next by thread: Re: disk latencie problems on a brand new exchange server with 2 mailboxes!!!!!!!!
- Index(es):
Relevant Pages
|
