Re: HELP: 80286 Informix 3.3 vs. SCO Unix/386 3.2.2

Path: emory!wupost!zaphod.mps.ohio-state.edu!think.com!hsdndev!rutgers!pyrnj!pyramid!infmx!aland From: aland@informix.com (Colonel Panic) Newsgroups: comp.unix.xenix.sco,comp.unix.sysv386,comp.databases.informix Message-ID: <1991Nov2.221616.6478@informix.com> Date: 2 Nov 91 22:16:16 GMT References: <2079@tiamat.fsc.com> Sender: news@informix.com (Usenet News) Organization: International Brotherhood of Geeks, Local 619 In article <2079@tiamat.fsc.com> jim@tiamat.fsc.com ( IT Manager) writes: >I've tried to use a narrower aim in getting my current situation resolved, >but I've had little success. Perhaps a wider audience will produce a >greater possibility for a solution. > >Here's the situation: > >We have an accounting application written in Informix 3.3 (Informix's >original database product, NOT the SQL product). We started out using this >application with an 80286 version of Informix 3.3 on an Altos 2086 box. >Later, we put the accounting dept. on their own 386 PC, and started >using Xenix 386 as the operating system. In both of these configurations >(i.e. an 80286 and an 80386 version of the Xenix OS), the 80286 version of >Informix 3.3 ran just fine. > >Recently, we decided to install SCO Unix 386 (via ODT 1.1) on the >accounting computer. Since SCO Unix 386 listed Xenix-compatibility (even >for 286 Xenix binaries) as one of its attributes, I figured we were in good >shape. > >Well, under SCO Unix 386, the 80286 Informix 3.3 binaries do work. BUT, >they are very, very, very, very slow. For instance, some operations >(especially update operations) that used to take 5 minutes may now >take as long as 40 to 50 minutes. I'm not sure if it is the OS or the >Informix binaries that are at fault. If performance was acceptable under 386 XENIX, then the binaries would appear to be OK. Perhaps SCO UNIX's XENIX compatibility was targeted for accuracy rather than performance; can't blame them for that. Perhaps an SCO insider might have an explanation for the performance hit... Check what locking method your Informix 3.3 uses: if fcntl/flock (aka "sys5lock"), perhaps the SCO UNIX kernel has to go thru hoops to handle XENIX-based fcntl calls... >Here's the questions: > >1) Is there a 80386 version of Informix 3.3 out there somewhere? My > Informix sales rep doesn't seem to be able to find one in their > product lists. I'd imagine that if the problem is with the 80286 > binaries running slowly, that any 80386 version (e.g. one for Xenix 386, > ISC Unix 386, Intel Unix 386, etc) might help. Apparently not; I checked the machines list too. I think the version you are running may predate XENIX 386 (!) >2) Could the problem be in the way SCO Unix 386 deals with 80286 binaries > and is there any way to tune the system so these old binaries will > run faster? Out of my league... >3) Any other suggestions (except: repalce the application, or rewrite the > application in Informix SQL) that might be useful in keeping both the > Informix 3.3 applications and the SCO Unix 386 OS on the same box. Your best bet will be to convert the app to newer product. If it's just Ace and PERFORM stuff, conversion is pretty painless. If you have ALL-2 code, that's trickier. There is a migration tool called ESQL/C-ALL that can be helpful in migrating from Informix 3.3-based ALL, but you'll still need to move to ESQL/C or 4GL eventually, as ESQL/C-ALL is on its way out, too. >Thanks for any help. I sure hope someone out there has an answer for me. If this is adequate to be considered an "answer"... >James B. O'Connor jim@tiamat.fsc.com >Information Technology Manager voice 615/821-4022 x651 >Ahlstrom Filtration, Inc. fax 615/821-0616 Followups to comp.databases.informix. -- Alan Denney aland@informix.com {pyramid|uunet}!infmx!aland "In the cafeteria just after lunch, (well, not *just* after, more like *during* lunch, about 12:28; say 12:30, give or take a few minutes), I leaned back in my chair (it was one of those aluminum chairs, good strength-to-weight, like titanium but not quite; but then of course titanium would be a bit of an overkill). Anyway, I heard one of the girls talking about how boring she thought engineers could be."