home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!LOBBY.TI.COM!RSNYDER
- X-Mailer: ELM [version 2.2 PL16]
- Message-ID: <9207241418.AA07940@ti.com>
- Newsgroups: bit.listserv.sas-l
- Date: Fri, 24 Jul 1992 09:18:17 CDT
- Reply-To: "R. Snyder" <rsnyder@LOBBY.TI.COM>
- Sender: "SAS(r) Discussion" <SAS-L@UGA.BITNET>
- From: "R. Snyder" <rsnyder@LOBBY.TI.COM>
- Subject: Re: Using SAS for CPE - related question
- Comments: To: MJF@IBM-B.RUTHERFORD.AC.UK
- Comments: cc: sas-l@uga.cc.uga.edu
- In-Reply-To: <9207241343.AA06363@ti.com>; from "Mike Froggatt" at Jul 24,
- 92 2:28 pm
- Lines: 40
-
- > o-o-o-o-o-o-o-o-o-o-o-o
- >
- > I am concerned about our current techniques for presenting
- > statistics on batch job turnround information for our mainframe
- > services. Currently we take a fairly simple approach and quote
- > things like:
- >
- > average time spent waiting in queues by job class
- > average expansion factors (that is, runtime/cpu)
- > 80th percentiles for queue, run times
- > (ie 80% of total CPU was done by jobs having an expansion factor <=6)
- >
- > The advantage of such numbers are that they are easy to quote and understand
- > to our management and users.
- >
- > The disadvantage is that they can also be very misleading, averages
- > are simple but do not give indications to variability in the underlying
- > data. A suggestion made is to parameterise the data and follow trends
- > based on these statistics.
- >
- > My question is: can anyone pass on useful techiques that they use to
- > report on their job turnround or give pointers to references on the topic.
- >
- > For the record, our performance data for our mainframe services
- > is in SAS under VM/CMS
- >
- > Any suggestions greatly appreciated.
- >
- > Mike Froggatt, Marketing Services Group
- > Rutherford Appleton Laboratory
- > Didcot, Oxford, England
- >
-
- Mike,
-
- I don't know much about CPE but if you can get your data into SAS data sets
- you should be able to use the Shewhart procedure in SAS/QC to present your
- data in an understandable way while showing variation and its significance.
-
- Bob Snyder, Texas Instruments, Sherman, TX
-