[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20050304: ADDE performance

Hi,
We have been looking at the SERVER.LOG files in an effort to
optimize our McIDAS servers. I got some old programs to read
out its contents. -chkdde.c dde.c servacct.h-

Inside servacct.h, CPU times come as:
long int cpu;     /* 100 X cpu seconds */

When I add up all McIDAS transactions on a busy server I get
values round 1000 ~ 2000 secs per day (15 ~ 30 minutes)
I never compared before these kind of figures. Any of you ever
played around adding up CPU secs for a McIDAS server?
In that case, what value ranges should I use to evaluate performance?

To the LDM forum: Any experience on getting results for ADDE
similar to what rtstats gives for LDM?

Pepo
+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+
Jose Juega                         Instituto Nacional de Meteorologia
Tecnico de Sistemas                          Leonardo Prieto Castro,8
Area de Telematica      JJJJJJJ   JJJJJJJ              E-20071 Madrid
I.N.M.-Madrid-SPAIN       JJJ       JJJ          Ciudad Universitaria
UAM (McIDAS)              JJJ       JJJ     Tel :     +34 91 581-9654
------------          JJJ JJJ   JJJ JJJ     FAX :     +34 91 544-5307
____________           JJJJJ     JJJJJ    e-mail:         address@hidden
+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+*+
Visit http://www.geocities.com/pepoj


#############################################################
This message is sent to you because you are subscribed to
 the mailing list <address@hidden>.
To unsubscribe, E-mail to: <address@hidden>
To switch to the DIGEST mode, E-mail to <address@hidden>
To switch to the INDEX mode, E-mail to <address@hidden>
Send administrative queries to  <address@hidden>

Return-Path: address@hidden
Received: from ssec.wisc.edu (mahogany.ssec.wisc.edu [128.104.110.2])
        by unidata.ucar.edu (UCAR/Unidata) with ESMTP id j24CBsv2019741;
        Fri, 4 Mar 2005 05:11:54 -0700 (MST)
Organization: UCAR/Unidata
Keywords: 200503041211.j24CBsv2019741
X-ListServer: CommuniGate Pro LIST 4.2.9
List-Unsubscribe: <mailto:address@hidden>
List-ID: <mcidas-users.ssec.wisc.edu>
List-Archive: <https://ssec.wisc.edu:443/Lists/mcidas-users/List.html>
Message-ID: <address@hidden>
Reply-To: "McIDAS Users list" <address@hidden>
Sender: "McIDAS Users list" <address@hidden>
To: "McIDAS Users list" <address@hidden>
Precedence: list
Date: Fri, 4 Mar 2005 06:11:46 -0600 (CST)
From: "John Benson" <address@hidden>
X-Original-Message-ID: <address@hidden>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Subject: Re: ADDE performance


    CPU used to be very important because it was often the
limiting resource.  In today's systems, the balance has changed
so much that most bottlenecks are elsewhere: memory, disk, comm.
30 minutes of CPU a day (out of 1440) is about 2%, far from
a sign of danger.  Of course, requests for meterological data
are usually peaky, so the short term load could be much greater.
Does vmstat or top indicate CPU saturation?  If so, is it due
to server activity, or something else?  Look for disk busy.  Look
for paging.  Look for network congestion.

    The logs will show you elapsed time as well as CPU time.
The ratio between them indicates the extent to which processor
time is effecting the total transaction time.  Fluctuations in
this ratio indicate how much you're being affected by other
factors.

    While it's sensible to use the logs as a tuning tool, that
isn't the reason they were implemented.  It was for billing.  And
current billing implementations care more about total bytes transferred
than they do about CPU.

   --johnb

#############################################################
This message is sent to you because you are subscribed to
 the mailing list <address@hidden>.
To unsubscribe, E-mail to: <address@hidden>
To switch to the DIGEST mode, E-mail to <address@hidden>
To switch to the INDEX mode, E-mail to <address@hidden>
Send administrative queries to  <address@hidden>