RE: oracle parallel server

Hi Chuck,

        We are using OPS on 2 node IBM HACMP's with shared SSA.

        OPS reduces performance but improves availability.

	I don't belive in the 5 9's bs going around as if you are 
	properly redundant everywhere and you have an unscheduled
	outage it is likely to be catastrophic. As I say to my managers,
	if I am down at all, it is going to be a minimum of 2 -3 hours to
	get back up. ( I call them managers instead of damagers because
	they actually listen)

	My record for 100% uptime (database availabilty) is 17 months
	on a standalone 7.2.2.3 database on a standalone SUN 5000. I
	guess that was before this 248 day problem.

	Currently we are not sophisticated enough here to achieve 
	this level with OPS. (Development and support issues)

	The big thing is everything required to provide access to the 
	database must be redundant at the hardware level. Power, network, 
	disks, controllers, machines, routers, switches, everything. 
	This makes it very expensive to implement properly.

	If you have the proper infrastructure and people the only downtime
	you should have is for OS upgrades and Oracle upgrades (patches can
	be applied one node at a time). 

	Unless your company is a finance institution or managing life and death data
	I would recommend avoiding OPS. Set up a hot standby or replicated site
	and use the money you save to provide more CPU/memory/disk for Oracle.

	I have seen more downtime because of inadequate hardware resources in 
	my career than because of disasters or hardware failures.

HTH
Dave

-- 
Dave Morgan
Senior Database Administrator
Internet Barter Inc.