Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!ginosko!uunet!ncrlnk!ncrwic!encad!entec!jlohmeye From: jlohmeye@entec.Wichita.NCR.COM (John Lohmeyer) Newsgroups: comp.periphs Subject: Re: SCSI->SCSI vs. SCSI->ESDI disk performance? Keywords: SCSI ESDI MD21 Sun-3 disk performance Message-ID: <374@entec.Wichita.NCR.COM> Date: 21 Jul 89 14:03:50 GMT References: <398@sunny.ucdavis.edu> Reply-To: jlohmeye@entec.UUCP (John Lohmeyer) Organization: NCR Corporation, Wichita, KS Lines: 37 In article <398@sunny.ucdavis.edu> poage@sunny.ucdavis.edu (Tom Poage) writes: >I have run into an apparently unresolvable question. >Is there a performance hit in having an ESDI disk >hooked to a SCSI bus through a translator (Emulex MD21) >versus a SCSI disk hanging directly on the bus? > You ask a very difficult question. So much depends on factors that you did not specify. Knowing the system and the controller helps, but not knowing the application probably would render any answer invalid. I say 'probably' because it may be the case that any application would be affected in an adverse way by adding the MD21. But I doubt it. Many SCSI to ESDI controllers actually delivery better performance than an embedded SCSI controller. In fact, it is possible to see better performance than the raw performance specifications of the ESDI drive would seem to indicate. Especially if the MD21 is a caching controller (I am not familiar with it). There is really only one good way to compare performance of two configurations. You have got to benchmark both configurations with a REALISTIC benchmark program. A realistic benchmark is one that accesses the disk in a pattern that is similar to the application you intent to run. In fact, the most realistic benchmark IS the application you intent to run. Although it is often difficult to get objective data from a real application. Unfortunately, our industry has a number of standard benchmark programs that often are poor predictors of actual system performance. I have limited first-hand experience with these, but have heard horror stories. Vendors then often fall into the trap of tuning their systems to run the benchmarks well. The benchmarks then run like a bat-out-of-hell, but the real applications just don't seem to do so well... -- John Lohmeyer J.Lohmeyer@Wichita.NCR.COM NCR Corp. uunet!ncrlnk!ncrwic!entec!jlohmeye 3718 N. Rock Rd. Voice: 316-636-8703 Wichita, KS 67226 SCSI BBS 316-636-8700 300/1200/2400 24 hours