Re: paging space size - contempoary best practice.
- From: "Mills, John T" <John.Mills@xxxxxxx>
- Date: Thu, 17 Jan 2008 08:44:34 -0600
Easy answer? 8GB
With todays drive sizes (vio excluded) 8GB is cheap and easy.
Hard answer?
Don't page. Buy more memory for servers that consistently page. Some
apps (messaging usually) will always page some because they are coded to
prefer swap pages for some work. For massive servers, say 128GB, that
consistently page 10GB or more, give them paging on dedicated drives in
the range of 64GB. Bleed hd6 to be 50% of actual paging in a normal
state. That way you can bleed beyond normal without a hang. Most don't
suggest this because you get a skip on the round robin when a space is
100%, but, since you don't hit 100% in hd6 until you are decidedly
abnormal, I wouldn't worry too much about it.
John T. Mills
________________________________
From: IBM AIX Discussion List [mailto:aix-l@xxxxxxxxxxxxx] On Behalf Of
Jackson, James D. (Mission Systems)
Sent: Wednesday, January 16, 2008 12:18 PM
To: aix-l@xxxxxxxxxxxxx
Subject: Re: paging space size - contempoary best practice.
I agree with the tuning approach suggested by Mr. Ray. It's difficult
to recommended a rule-of-thumb or best practice for sizing paging space
on new or upgraded hardware, as much depends on the type of database
(e.g., transactional vs data mine), real-time performance, filesystem
type (or use of raw devices), and VMM tuning. Unfortunately, I haven't
had the pleasure of working with these types of workloads, lately.
You may want to search the IBM Redbooks for more information relevant to
your specific software and hardware architectures.
http://publib-b.boulder.ibm.com/Redbooks.nsf/portals/
Regards,
James Jackson
________________________________
From: IBM AIX Discussion List [mailto:aix-l@xxxxxxxxxxxxx] On Behalf Of
Miller, Dave (I.S.)
Sent: Tuesday, January 15, 2008 9:36 AM
To: aix-l@xxxxxxxxxxxxx
Subject: paging space size - contempoary best practice.
Does anyone have the 'IBM best practice' for sizing your paging space on
large RAM enabled servers? I know in the past this question had many
varied answers. I'm curious if anyone knows of a specific IBM document
that is relatively recent that has a statement about it.
For example, we are upgrading two servers 32GB to 128GB. My past notes
and research would have me size paging space at 1.5*128=172GB....do
folks generally do this??? Is there a technical reason for it?
Ironically these are some of the very same servers that we strive to
keep utilization of paging space as low as possible...
Thanks for any input you can provide.
---------------------------------
David Miller
Database Architect
Baystate Health System
3601 Main Street, Springfield, MA 01109
413-794-3836, dave.miller@xxxxxxx <mailto:dave.miller@xxxxxxx>
------------------------------------------------------------------------
------------------------------------------------------------------------
-
Fools ignore complexity. Pragmatists suffer it. Experts avoid it.
Geniuses remove it (Alan J. Perlis)
------------------------------------------------------------------------
------------------------------------------------------------------------
-
- Prev by Date: Re: P570 and power regulators
- Next by Date: Re: Replacing rootvg mirror disk
- Previous by thread: Re: paging space size - contempoary best practice.
- Next by thread: darcs Binaries
- Index(es):
Relevant Pages
|
|