Re: Write efficiency

According to documentation:

Write efficiency
        Write efficiency is the difference between Write Commands (DSWRTS)
and Device Write Operations (DSDWOP)
        divided by Write Commands (DSWRTS), expressed as a percent.

This is a measure of disk cache efficiency - basically, how good device can
combine several write requests, received from system to one write operation
to device.
This is a function of cache implementation (which cannot be controlled),
but also how random disk accesses are.
In ideal case, if one job writes sequentially to a file - with a large
blocking factor - (which translates to a lot of writes to sequential
sectors on disk) this figure should go up. How close to 100% it can go, I
do not know, but I would estimate it as 60-70%.
In the worst case, when many jobs pound disk with many short random
requests, located in different areas of the drive, it will ten to go to
zero.
If you experience 40% write efficiency, I would estimate it is fairly good.
Nothing to worry about.

There's no special tuning, which specifcally improve cache efficiency.
Just good old guidelines - use sequential access with reasonably large
block size whenever possible; isolate heavily used objects in a separate
ASPs; periodically reload large files with high shrink-expand activity
(this will better distribute the file and reduce contention for every
single drive).


Best regards
    Alexey Pytel



john.lewis-crosby@oxinst.co.uk on 04/08/99 11:14:25 AM

Please respond to MIDRANGE-L@midrange.com

To:   MIDRANGE-L@midrange.com
cc:    (bcc: Alexei Pytel/Rochester/IBM)
Subject:  Write efficiency





I'd be grateful for some help in understanding the write efficiency
parameter in the Disk Activity section of the Performance Monitor Component
Report.  We have an S30, 1Gb main storage, 18 4Gb disks on 2 IOP's, running
a client/server ERP application.  Minimal green screen.
We are getting Write Efficiency figures per disk of 1 to 45%.   The single
average figure shown in the summary section below is typically 40%,
(overall disk utilization over the same period is 8 to 10%).

My question is what do these write efficiency figures mean in practical
terms, i.e.
How relevant/important are they?
Is 40% bad or good?
If bad, what effect are they having on overall system performance?
How can they be improved - system parameter tweaks, more Main Storage, more
IOP's, more disks?
Is there a "wall" at which performance falls off drastically?

We have had a scout through the manuals but would appreciate any pointers
to relevant sections.

Thanks, John Lewis-Crosby

************************************************
Head Office, Oxford Instruments plc
Old Station Way, Eynsham
Witney, Oxfordshire, OX8 1TL, England
Tel: +44 (0)1865 881437, Fax: +44 (0)1865 881944
URL http://www.oxford-instruments.com/
************************************************
+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to
MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator:
david@midrange.com
+---



+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---