home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!howland.reston.ans.net!zaphod.mps.ohio-state.edu!uwm.edu!psuvax1!psuvm!auvm!ESPRESSO.RT.CS.BOEING.COM!OLDENKMP
- Message-ID: <9301122116.AA02096@espresso.rt.cs.boeing.com>
- Newsgroups: bit.listserv.sas-l
- Date: Tue, 12 Jan 1993 13:16:31 PST
- Reply-To: "P. Oldenkamp 5-4481" <oldenkmp@ESPRESSO.RT.CS.BOEING.COM>
- Sender: "SAS(r) Discussion" <SAS-L@UGA.BITNET>
- From: "P. Oldenkamp 5-4481" <oldenkmp@ESPRESSO.RT.CS.BOEING.COM>
- Subject: Re: Comparison SQL to data step
- Comments: To: SAS-L@uga.cc.uga.edu, grant@CICERO.SPC.UCHICAGO.EDU
- Lines: 49
-
- ----------------------------------------------------------------------
- CONTENT: Question on interpretation of STIMER and FULLSTIMER stats
- REL/PLTF: 6.07TS203/SunOS
- E-ADDR: oldenkmp@ESPRESSO.boeing.com
- NAME: Paul OldenKamp
- PHONE: 206-865-4481
- ----------------------------------------------------------------------
- Grant Blank <grant@CICERO.SPC.UCHICAGO.EDU> compared matching two SAS datasets
- using a data step and proc SQL and found that SQL was much faster. He concluded
- that proc SQL was 25 times faster that the data step from the fact that the
- cpu time reported with the STIMER option was .31 with SQL and 7.94 with the
- data step. But the log contained:
- ...
- > NOTE: DATA statement used:
- > real time 19.919 seconds
- > cpu time 7.940 seconds
- >
- ...
- > NOTE: PROCEDURE SQL used:
- > real time 4.481 seconds
- > cpu time 0.310 seconds
- >
- My interpretation would be that SQL was about four times faster than the data
- step since I would argue along the FM(Unix Companion, p. 226) that elapse time
- is more significant that cpu time. What is the significance of 'real time'
- and 'cpu time' reported here and can someone offer an argument in favor of
- using cpu time?
-
- Along the same lines, when I turn on FULLSTIMER I get a report like this:
-
- NOTE: PROCEDURE GLM used:
- time: memory:
- real 11:29.123 page faults 14
- user cpu 1:58.570 page reclaims 154
- system cpu 45.430 seconds usage 880 K
- block I/O operations: context switches:
- input 14 voluntary 11639
- output 11659 involuntary 14787
-
- But the FM is not very helpful in interpreting the information. The
- description of FULLSTIMER I found is(UNIX Companion, p. 153):
-
- "The FULLSTIMER system options prints the following statistics: real time, user
- CPU time, system CPU time, and various other statistics depending on what
- machine you use."
-
- My questions are: Where can one RTFM on these measures of performance? Can
- anyone explain them for us? How do the reports on other machines differ
- from these that I get with SunOS?
-