Re: Write performance.
From: GMN (anonymous_at_discussions.microsoft.com)
Date: 02/22/05
- Next message: Yogish: "RE: Single-user mode (server)"
- Previous message: DC Gringo: "Maintenance plan reports"
- In reply to: Andrew J. Kelly: "Re: Write performance."
- Next in thread: Andrew J. Kelly: "Re: Write performance."
- Reply: Andrew J. Kelly: "Re: Write performance."
- Messages sorted by: [ date ] [ thread ]
Date: Tue, 22 Feb 2005 12:14:44 -0800
I know Raid5 is bad for writes. But I'm just concerned
that performance goes WAY down for writes if anything
else is going on. The admin admitted that the write
cache on this box was pretty small.
Can you tell me any other configuration parameters that I
can question them about? (Disk size is pretty small -
36G - and I bellieve rpm's is pretty good as well.) Or
is write cache the main consideration under these
circumstances?
>-----Original Message-----
>A lot has to do with how much write cache your SAN has
and how it is
>configured. It also has to do with the fact you are
using a RAID 5 which is
>very write intensive. But anytime you attempt to do two
large operations
>you have to share the disks and performance will be
decreased accordingly.
>That is why a properly configured SAN with the right
amount of cache is
>critical for performance. SAN's are not magical, they
need to be configured
>to handle the load jsut like anything else.
>
>--
>Andrew J. Kelly SQL MVP
>
>
>"GMN" <anonymous@discussions.microsoft.com> wrote in
message
>news:070801c5190d$d49b8b80$a601280a@phx.gbl...
>> We've got a SS 2000 system with two Raid 5 SAN (EMC)
>> systems. I have been doing copies from the old slower
>> drive to the newer but have had poor write performance.
>> Here is what I observed this morning with our admin:
>> while I did a copy (using Windows Explorer from old to
>> new) under various situations:
>>
>> --When there was nothing else running on the server,
>> copying 2 G took about 2 minutes which is reasonable.
>> --When the copy was done simultaneously with another
>> simple SELECT query, the admin observed that the Write
>> Mb/Sec dropped to about a tenth of the level it was
doing
>> previously.
>> --I ran a test later and found that the copy occurred
3-
>> 5X more slowly with a simple SELECT query running in
the
>> background.
>>
>> I find all this disturbing, because it seems to imply
>> that if you have anything else running on the server,
>> write performance will drastically fall off. Is this
>> really to be expected? Does Sql Server hog performance
>> on a server by any chance?
>>
>> (I am wondering if there is some issue with the Raid(s)
>> or the EMC regarding Writes.)
>>
>>
>>
>
>
>.
>
- Next message: Yogish: "RE: Single-user mode (server)"
- Previous message: DC Gringo: "Maintenance plan reports"
- In reply to: Andrew J. Kelly: "Re: Write performance."
- Next in thread: Andrew J. Kelly: "Re: Write performance."
- Reply: Andrew J. Kelly: "Re: Write performance."
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
