[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference ulysse::rdb_vms_competition

Title:	DEC Rdb against the World

Moderator:	HERON::GODFRIND

Created:	Fri Jun 12 1987
Last Modified:	Thu Feb 23 1995
Last Successful Update:	Fri Jun 06 1997
Number of topics:	1348
Total number of notes:	5438

1144.0. "Sybase performance" by AUNTB::KULVETE () Fri May 01 1992 16:26

    Here's the scoop at one of my customers.  They're looking at Sybase on
    an RS6000 model 530 (I believe) and on a VAXstation 4000 model 60.
    They have written a program that updates the same record 100 times.  It
    takes 28 seconds on the VAX and 1 second on the RS6000.  I would expect
    maybe a 4-6 times difference, but not this great.  Sybase is going to look
    into this, but I don't see much motivation for them to figure this out if
    my customer will buy Sybase on an RS6000.  This is the same database and
    the same data.  Apparently, the CPU on the VAX is getting eaten up.  A
    person out there also tried something similar with Rdb on the VAX and it
    took 3-4 seconds for the same number of updates (not the same databse
    though).
    
    Does anyone have any ideas as to why there is such a difference?
    
    This is my last day in sales support here in the Carolinas (I'm going to
    be up in ACMS engineering), so could you send any information, suggestions
    to either AUNTB::CLARK (Sandy - database sales support) or AUNTB::GILSTER
    (Bill - sales rep).
    
    Thanks.
    
    Dave

T.R	Title	User	Personal Name	Date	Lines
1144.1	Apples and oranges	BAHTAT::DODD	gone to Helen's land	`Fri May 01 1992 17:53`	10
	My first stab at this would be that Sybase on VAX/VMS is writing the data to disk for every update, Sybase on UNIX is writing to memory and only flushing at the end (or maybe not at all!). This is the usual way unix would be set up, writes are done to memory and periodically flushed. Is this good or bad - if the system stays up its good, if the system crashes give me the VAX every time. Caveat:- I don't know this for a fact but it would be the first thing I'd look into. Andrew
1144.2	Motivate them :-)	COOKIE::BERENSON	Lex mala, lex nulla	`Fri May 01 1992 19:21`	25
	re .1: Doubtful, unless SYBASE parameters are set incorectly on VMS. SYBASE uses the same buffer flush and recovery strategies on both operating systems. I don't know the raw CPU horsepower of the RS6000 in question but I imagine its better than the 4000-60. General industry trends are that CISC does MUCH better than RISC when comparing SPECint per TPS. So, one might expect the 4000-60 to equal or outperform the RS6000. SYBASE runs pretty well on VMS so I'm really surprised at the result. I'd have to guess that its either (1) bad tuning on the VAX or (2) a bug. Vis a vi your Rdb number, unless its an apples to apples comparison using a fully tuned Rdb then I wouldn't even mention your number to anyone. My guess is that Rdb on the 4000-60 would beat SYBASE on the RS6000 given Rdb V4.1 and proper tuning (and the identical application). Hal Ps: I concur with you that SYBASE has little incentive to help the VMS sale when they win either way. Still, given the magnitude of the difference SYBASE is probably worried. For example, if you told SYBASE that they were running this application an order of magnitude slower than Rdb/VMS on the 4000-60 they might be scared we could use that result against them elsewhere. That would motivate them to help solve your customer situation and take away a competitive black eye for them.