home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!SUVM.BITNET!SASMAINT
- Return-Path: <@OHSTVMA.ACS.OHIO-STATE.EDU:SAS-L@VTVM2.BITNET>
- Return-Path: <@AWIIMC12.IMC.UNIVIE.AC.AT:SASMAINT@SUVM.BITNET>
- Message-ID: <SAS-L%93012523232232@AWIIMC12.IMC.UNIVIE.AC.AT>
- Newsgroups: bit.listserv.sas-l
- Date: Mon, 25 Jan 1993 16:34:06 EST
- Reply-To: "Nelson R. Pardee" <SASMAINT@SUVM.BITNET>
- Sender: "SAS(r) Discussion" <SAS-L@UGA.BITNET>
- From: "Nelson R. Pardee" <SASMAINT@SUVM.BITNET>
- Subject: SAS Performance, Hard Disk controllers
- Lines: 53
-
- A SUGI plug! We're planning a repeat of last year's popular panel on
- deciding which platform to run on. Returning from last year by
- popular demand will be Phil Miller, John Sims, and Randy Betancourt,
- with your's truly moderating. Hopefully, this year it won't be late
- Wednesday morning!
-
- Performance issues are really tricky. There have been a couple of trade rag
- articles recently focussing on comparisons of running the same
- applications across platforms (specifically Windows, Mac, and some
- UNIX flavor). Interestingly enough, Windows platforms tend to come
- out ahead (I believe also over the OS/2 application in one instance).
- In one case, the reviewer was Windows biased but in the other the
- reviewer was workstation/Mac biased. These applications weren't SAS:
- they were word processing, graphics programs, and spreadsheets. So
- the point an earlier poster made about what other programs you run
- besides SAS is an important one. Although it's nice to have the
- fastest platform available, the platform only has to be fast enough
- for *your* use. The availability and cost of other applications may
- have a large say in deciding what is the optimal platform. It's your
- application that counts when it comes to deciding. ªone anomaly I
- found in benchmarking my application is that the unix boxes were
- fastest if I made certain I turned autoscroll off; if on, they
- couldn't beat a PC- I had a really large programß.
-
- Hard disk controllers: there is some literature that indicates that a
- good software cache is close to equivalent to a hardware caching
- controller. This seems to run counter to a previous posting; your
- mileage may vary. It's also not easy to know what caching is going
- on in your hardware; some cache is almost always included. I *think*
- many IDE drives included in Gateway systems have had some cache. One
- disadvantage I found to the EISA box (with a caching controller) is
- that the EISA machine is harder to maintain and slower to boot (a
- minor annoyance, but a pain when setting up new things, rebooting
- often). My gut feeling is that for most folks the extra $$ (up to
- 1000) premium EISA often commands is not worth it.
-
- Be sure when using either a hardware or software cache to set the
- number of buffers in config.sys to some small number. (If you're
- running Windows you're probably running a software cache- SMARTDRV).
- I've heard lots of recommendations, but I like 5 or fewer (I think
- Microsoft says 10). Be warned that lots of installation programs for
- various program will make arbitrary decisions and set it to some
- higher value without asking. So, you may have to check it on a
- regular basis.
-
- I could ramble on about local bus vs EISA, etc., but it's really
- beyond the scope of this list. PC magazine had a discussion that is
- much better than anything I can do in a sidebar with a recent
- article- I think a review of Video boards or PCs (sorry I don't have
- exact issue).
- --Nelson R. Pardee, Computing Services, Syracuse University --
- --120 Hinds Hall, Syracuse, NY 13244-1190 USA (315)443-1079 --
- --Bitnet: SASMAINT@SUVM Internet: SASMAINT@SUVM.ACS.SYR.EDU --
-