Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> RE: UNIX nice Command Good/Bad

RE: UNIX nice Command Good/Bad

From: DENNIS WILLIAMS <DWILLIAMS_at_LIFETOUCH.COM>
Date: Thu, 08 Aug 2002 13:27:17 -0800
Message-ID: <F001.004AFCF9.20020808132717@fatcity.com>


Anjo - The way that I would put it is that the operating system (any multitasking o.s.) must make a decision on which process to give which processor next and for how long. It is more complex than the o.s. just picking the highest priority process from the run queue (processes that are available to run, not waiting for I/O for example). Each o.s. vendor has its own proprietary decision process. The o.s. lowering the priority of a process that has accumulated a lot of CPU time is one factor. The priority figure is one way to make provide some visibility of a single aspect of this decision process. Tru64 tracks the "nice" value in addition to priority. Some vendors do better at this than others. One sign of doing a better job is being able to scale to more processors.

   I took a C.S. class in operating systems years ago and the main thing I came away with was:

1. It isn't simple.
2. Most vendors configure their system pretty well as delivered.
3. It is better to change the priority factors in your application before
you try changing the operating system's parameters. 4. If you decide to change something, just change one thing at a time and measure the effect.

This is as much as I can recall today.
Dennis Williams
DBA
Lifetouch, Inc.
dwilliams_at_lifetouch.com

-----Original Message-----
Sent: Thursday, August 08, 2002 3:26 PM
To: Multiple recipients of list ORACLE-L

Not sure if this has been mentioned some where, but it used to be that UNIX would lower priority of processes that had accumulated a lot of CPU time. So the Oracle Background processes would get lower priority over time. Stopping and starting an instance will fix this again ;-)

Cary Millsap wrote:

>Anything that an OS does to raise or lower process priority without the
>Oracle kernel's knowing is a bad thing.
>
>Example: An inefficient process called L that makes way too many cache
>buffer chains acquisitions is demoted ("nice"d to a less important
>priority). An important job called H that needs some of the same latches
>is promoted ("nice"d to a more important priority). L, which gets some
>user-mode CPU time in spite of its poor priority, works long enough to
>acquire a latch, and then it gets preempted. Now, when H needs the same
>latch that L is holding, H will consume lots of CPU spinning for the
>latch, but of course H can never obtain the latch until L releases it.
>But L is asleep most of the time because it has a low priority. You'll
>have a busy-looking system (high CPU utilization because of H's
>spinning), but nobody will be able to get anything done.
>
>When the Oracle kernel provides process prioritization or suspend/resume
>capabilities, it'll be fantastic (I think they advertise some of this
>stuff in 9i, but I'm not sure it works yet). But no OS can do the job
>unless it takes into account some of the things that Oracle is doing
>internally.
>
>
>Cary Millsap
>Hotsos Enterprises, Ltd.
>http://www.hotsos.com
>
>Upcoming events:
>- Hotsos Clinic, Oct 1-3 San Francisco, Oct 15-17 Dallas, Dec 9-11
>Honolulu
>- 2003 Hotsos Symposium on OracleR System Performance, Feb 9-12 Dallas
>- Next event: NCOAUG Training Day, Aug 16 Chicago
>
>
>
>-----Original Message-----
>Ethan
>Sent: Tuesday, August 06, 2002 5:23 PM
>To: Multiple recipients of list ORACLE-L
>
>There is an article on Metalink that talks about doing this and doesn't
>mention the problem but I will take your word over that. Another post
>on
>usenet talked about some different software available on particular
>operating systems that would do the smae thing as "nice" but I guess it
>does
>is a little more "nicley". Would this be safe? How about punishing the
>whole lot and forcing all processes with a particular database to run at
>a
>lower level?
>
>Ethan Post
>(972) 577-6552
>Ethan.Post_at_ps.net
>perotdba (AIM), epost1 (Yahoo)
>--------------------------------------------------------------------
>
>
>-----Original Message-----
>Sent: Tuesday, August 06, 2002 4:28 PM
>To: Multiple recipients of list ORACLE-L
>
>
>Still true.
>
>
>Cary Millsap
>Hotsos Enterprises, Ltd.
>http://www.hotsos.com
>
>Upcoming events:
>- Hotsos Clinic, Oct 1-3 San Francisco, Oct 15-17 Dallas, Dec 9-11
>Honolulu
>- 2003 Hotsos Symposium on OracleR System Performance, Feb 9-12 Dallas
>- Next event: NCOAUG Training Day, Aug 16 Chicago
>
>
>
>-----Original Message-----
>Ethan
>Sent: Tuesday, August 06, 2002 3:58 PM
>To: Multiple recipients of list ORACLE-L
>
>Is this still true?
>
>************************************************************************
>****
>*
>WARNING: do not run Oracle processes at different priority by altering
>the nice value. This is specifically warned against in the manuals.
>If you have a process with a lowered priority that obtains a lock on
>an object, then can't run because of the low priority, you will create
>a deadlock or hang situation. All Oracle processes should run at the
>same priority.
>
>

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: Anjo Kolk
  INET: anjo_at_oraperf.com

Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).
-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: DENNIS WILLIAMS
  INET: DWILLIAMS_at_LIFETOUCH.COM

Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).
Received on Thu Aug 08 2002 - 16:27:17 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US