[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

1133.0. "Significant increase in parameter settings" by MJBOOT::SCOTT () Wed Apr 08 1992 23:04

    Hi,
    
    
    I am not an Oracle or RDB heavy but I am currently a resident at a
    large VAX site that runs ORACLE. They are about to upgrade their ORACLE
    software from version 6.032 to 6.2 and have requested the following 
    sysgen parameters to be significantly increased ....
    
       		New Values	Current    Maximum VMS Setting
    
    SRPCOUNT        175000        17000		270336
    SRPCOUNTV       275000        85000         270336
    IRPCOUNT         60000         8960         135168
    IRPCOUNTV       200000        35840         135168
    LOCKIDTBL       175000        24000         262144
    LOCKIDTBL_MAX   275000        65535         262144
    REHASHTBL         8192   	   8192           8192
     
    Several of these meet or exceed the allowable maximum for VMS V5.4.
    Are these for real or should I question these values with the on-site
    Oracle rep who requested them.
    
    Do they really need these or are they blowing smoke ?
    
    
    Thanks,
    
    Tim

T.R	Title	User	Personal Name	Date	Lines
1133.1	If they use the DLM they'll need some increasing	COOKIE::OAKEY	Picard/Riker '92	`Wed Apr 08 1992 23:14`	14
	� <<< Note 1133.0 by MJBOOT::SCOTT >>> � -< Significant increase in parameter settings >- Tim, I don't have a clue about Oracle, but these parameters do deal lock management in a distributed (aka VAXcluster) environment. If Oracle 6.2 uses the VMS lock manager, then increasing these parameters isn't surprising to me (Rdb also requires well sized lock-related parameters). I'd check to see if 6.2 uses the lock manager (ie, $ENQ and $DEQ system services) and if the size is accurate if they do use the lock manager, but in using the lock manager I wouldn't really question the fact that they need to be increased...
1133.2	Parallel Server ?	KCOHUB::DAZOFF::DUNCAN	Gerry Duncan @KCO, 452-3445	`Thu Apr 09 1992 16:55`	7
	We need to know if this customer is going to use the cluster capability of Oracle referenced in -.1 AND if the customer is going to implement Oracle's parallel server. Also, can someone post the Rdb default sysgen parameters ? -- gerry
1133.3	Parallel Server & Cluster Capability	MJBOOT::SCOTT		`Thu Apr 09 1992 18:15`	12
	< We need to know if this customer is going to use the cluster capability < of Oracle referenced in -.1 AND if the customer is going to implement < Oracle's parallel server. Yes and yes. Tim
1133.4	those number are HUGE	COPCLU::BRUNSGAARD	Curriculum Vitae, who's that ??	`Fri Apr 10 1992 13:58`	33
	But these parameters are ENORMOUS !!! I personnaly haven't anything like this in any Rdb installation, but maybe mine are just just small -:) The largest being a 1000 terminal ACMS/RDB based system using 40 Acms Servers in peak periods. The settings are (of the top of my head) SRPCOUNT < 100000 Srpcounv doesn't matter too much since it is only the ultimate limit IRPCOUNT < 40000 ...v same as above Lockidtbl < 80000 Reshhashtbl always set to 1/4 of lockidtbl. I would claim that running with a lockidtbl of 275000 and reshashtbl of 8192 is plain nuts !! Note that SRPCOUNT is charged directly against your physical memory on your machine, and these numbers are HUGE. My guess is that something is fishy - either the settings are wrong - or Oracle REALLY uses ressources I don't know which is the right one (or a combination of the two). Anyway one thing is certain: Those settings can't be the default settings, so comparing them with Rdb default settings are not apples to apples comparison at all. Just an opinion lars