tunable kernel parameters for SCSI H.D.

tunable kernel parameters for SCSI H.D.

Post by Timothy D. Kue » Thu, 21 Sep 1995 04:00:00





Quote:>My company has a 486-66 running SCO Unix 3.2.4.1 with the following:
>  16MB RAM
>  1542B Adaptec SCSI Controller
>  3 1GB Micropolis SCSI Hard Disk
>  525MB Archive SCSI Tape Drive
>We are using RM/COBOL to store data.  The files are around 300MB,
>storing sales information.  Our problem is that whenever someone
>does a report covering a long period of time (which means a lot of
>disk I/O), the system slows down a lot.  A simple login will take
>about 30 seconds.

My first advice would be to add more memory. If SCO can keep the report
stuff in memory then you won't have to wait for the system to flush it's
buffers to get a copy of /bin/login before letting you on.

Another alternative would be to swap the ISA MB for a PCI MB.  This
alone will make a big difference in speed w/out going for more mem.

Quote:>Is there any kernel parameters for SCSI H.D. so that disk I/O won't
>be this slow.  I have looked at the books, and AIO doesn't seem to
>apply.

The disk I/O isn't that slow - you're stuffing more down it's throat
with those big reports than it can swallow at one time, which is why
it's taking so long for other things to get done.

One thought that comes to mind would be using nice(C) to lower the
run priority on your report so that other activity will more readily
pre-empt it. The report will take longer to run, but your users won't
suffer as much.

 -------------------------------------------------------------------

 -------------------------------------------------------------------

 
 
 

tunable kernel parameters for SCSI H.D.

Post by Jeff Lieberma » Sun, 24 Sep 1995 04:00:00



: Is there any kernel parameters for SCSI H.D. so that disk I/O won't
: be this slow.  I have looked at the books, and AIO doesn't seem to
: apply.

Yep.  However, there's not enough information here to make any
usable suggestions.  Your 1542B controller is relatively slow
compared to todays EISA and PCI controllers.  Your drives
may not be SCSI II/FAST (10Mbits/sec) and are stuck at 5Mb/sec.
Lots of places where things run slow without having the
Unix kernel mess things up.  Before you attack, buy a copy of:
        "The SCO Performance Tuning Handbook"
        by Miscovich and Simons.
Unfortunately, the book does NOT discuss many harware issues.
Replacing hardware will yield much greater improvements than
tweaking kernel parameters.

I'll assume that you don't want to change hardware.  In
3.2v4.x, SCO does one thing wrong.  NBUF (disk buffers)
is limited to 600KB in the stock configuration.  This is
much too small for a typical 16MB ram system.  Try putting
about 1/3 of your memory into disk buffering.  Run:
        cd /etc/conf/cf.d
        ./configure
and set:
        NBUF  = 4096     (4MB of disk buffers)
        NHBUF = 1024     (1MB of hash buffers)
Relink the kernel and reboot.  See if that helps.

If your Cobol application generates lots of temporary files
and thrashes around the disk a bunch (like the one's I've
seen), then a large disk cache will help considerably.  Also,
make sure you have a good SPS power system.  Leaving 4MB of
live data in memory after a crash is a great way of
destroying the file system.

Additional commands to try are:
        u386mon     (version 2.7)       (System monitor)
        /usr/bin/sar                    (System activity report)
        /usr/bin/uptime                 (Shows current load averages)
        /etc/conf/cf.d/configure -x     (list the kernel parameters)
        /etc/sysdef                     (display tunable parameters)
You will need these to determine if you are making any progress.

There are also commercial products that will help with tuning.

        SarCheck by Aurora Software (603)382-9264

        Monitor by Stallion Tech.

This is a VERY superficial coverage of a non-trivial subject.
Good luck.

--
# Jeff Liebermann  Liebermann Design  150 Felker St #D  Santa Cruz  CA  95060

# 408.699.0483 digital_pager    73557,2074  cis [don't]


 
 
 

tunable kernel parameters for SCSI H.D.

Post by Greg Haye » Tue, 26 Sep 1995 04:00:00




> Path: hayford.demon.co.uk!news.demon.co.uk!dispatch.news.demon.net!demon!
> tank.news.pipex.net!pipex!news.mathworks.com!uunet!in1.uu.net!
> news.sprintlink.net!news.uoregon.edu!serv.hinet.net!usenet

> Newsgroups: comp.unix.sco.misc
> Subject: tunable kernel parameters for SCSI H.D.
> Date: 20 Sep 1995 16:01:39 GMT
> Organization: HiNet
> Lines: 19

> NNTP-Posting-Host: 168.95.231.188

> My company has a 486-66 running SCO Unix 3.2.4.1 with the following:
>   16MB RAM
>   1542B Adaptec SCSI Controller
>   3 1GB Micropolis SCSI Hard Disk
>   525MB Archive SCSI Tape Drive

> We are using RM/COBOL to store data.  The files are around 300MB,
> storing sales information.  Our problem is that whenever someone
> does a report covering a long period of time (which means a lot of
> disk I/O), the system slows down a lot.  A simple login will take
> about 30 seconds.

> Is there any kernel parameters for SCSI H.D. so that disk I/O won't
> be this slow.  I have looked at the books, and AIO doesn't seem to
> apply.

> Thanks in advance for any suggestions.

We also use RM/Cobol to do heavy processing. I would use sar to
determine where your computer resources are going. RM/Cobol is very
hungry on CPU usage.  We also use LPI/Cobol for some processes. The
advantage is that it produces machine code and is not an interpreter
like RM/Cobol.  For most applications it is 40 to 100 times as fast
in straight CPU usage. Overall programs can run 20 times as fast depending
on the ratio of I/O to CPU. The only advantage of RM is that the Indexed
Sequential file access can hold the data and indexes in a compressed form,
taking a lot less space than LPI. If you want anymore info then mail me.
--

Greg Hayes
MD
Hayford Systems Ltd
+44 71 247 3005 Tel
+44 71 247 3280 Fax
 
 
 

1. SCO Unix 3.2.4.2 tunable kernel parameter for SCSI H.D.


: My company has a 486-66 with 16MB RAM, Adaptec 1542B SCSI controller
: card, and 3 Micropolis 1GB H.D.  Everytime someone does a report, the
: system slows down quite a bit (it was due to that RM/COBOL data files
: are about 50MB in size, and sometimes the report will cover 3 years of
: data).  Is there any tunable kernel parameter that can adjust the
: behavior of the SCSI H.D. disk access?

What version of SCO Unix?

There is no way you are going to cache a signifigant part of
a 50Megabloat data file in 16MB of ram.  Subtracting about 4MB
for the kernel and 0.5MB per user, there really isn't that much
ram left to build a decent cache.  You *probably* need to add
some ram.  (This is a guess, this is only a guess).

If you're running 3.2v5.0.x, the NBUF and NHBUF kernel disk buffer
parameters are established dynamically.  Not much you can do there.

However, if you are running 3.2v4.2, SCO sets a limit of 600Kb of
NBUF disk buffering for boot time buffer allocation.  IMHO, this
is much too small.  I usually set NBUF for about 1/4 to 1/3 of real
memory with NHBUF=NBUF/4 rounded up to the nearest power of 2.
Several of my customers machines (one running RM/Cobol) have
achieved dramatic improvements in disk performance.  It doesn't
do anything to sequential access (large file copies and tape backups)
but random access is much better.

For 3.2v4.2, I suggest "The SCO Performance Tuning Handbook" by
Miscovich and Simons, ISBN 0-13-102690-9.  You will also need
to monitor sar output to determine if you're making any progress.

[x] Email to author  [ ] To mailing list  [x] Posted to newsgroup
--
# Jeff Liebermann  Liebermann Design  150 Felker St #D  Santa Cruz  CA  95060

# 408.699.0483 digital_pager    73557,2074  cis [don't]

2. Where can I Find FDISK ?

3. Perl Module Installation Error

4. OSR5.0.0 -- tuneable kernel parameters

5. General Protection Fault

6. New: Change compile-time option CONFIG_APM_RTS_IS_GMT to sysctl-tunable kernel parameter

7. NASA Computer Security Conference

8. Tunable Kernel Parameters

9. List of Tunable kernel parameters

10. Solaris 2.3 tunable kernel parameters

11. Kernel TCP/IP Tunable parameters?

12. kernel tuneable parameters?