Re: ASM for single-instance 11g db server?
Date: Thu, 07 Apr 2011 23:31:17 +1000
Message-ID: <inkeb0$9tm$1_at_dont-email.me>
joel garry wrote,on my timestamp of 7/04/2011 3:21 AM:
>
> Of course if one of many controllers corrupts, welcome to hell.
Exactly. RAID5, RAID10, whatever, it doesn't matter. That is why we have backups and DR sites. No RAID type was ever thought out to prevent ALL possible disaster scenarios.
> Never underestimate the power of the damager:
> http://groups.google.com/group/comp.databases.oracle.server/msg/1cedb062e0071ac0
> http://groups.google.com/group/comp.databases.oracle/msg/29f1684073028648
Dude: the last time one damager walked into our DC we made sure he couldn't walk ever again!
> I've been amazed many times how often warnings are ignored. (Couldn't
> find with a quick search something I came across yesterday on the hp
> forums - guy posted his warning messages with a raid question,
> apparently didn't realize that the "battery dead" messages meant no
> battery backup on write cache [depends on hardware as to whether that
> turns cache off, I believe].) Yet another use for duct tape -
> covering up annoying warning lights.
> http://groups.google.com/group/comp.databases.oracle.server/msg/9be05efc4b33ac4c
Sure. But like I said: RAID was not designed to cover ALL possible disaster scenarios, natural or stupidity-caused.
> I'm running OLTP on RAID5. No one mentions what happens when the
> write buffer is full. Or when the batteries die and turn off the
> cache.
That is indeed another case of "traditional lore" creeping in.
Not all OLTP workloads are the same. I ran a OLTP system in my search engine marketing days for two years with 0 (zero) downtime, using 9i and RAID5 on an Apple XRaid SAN: can't get any more low end! Yet the workload was of a nature that allowed us to get away with it: very rapid arrival of very short transactions, but not overlapping or multi-threaded. We were happily pumping an average of 100MB/s, constant 24X7, at a ridiculously low cost. 6 years ago. No problems.
When it comes to storage performance and resilience, there is no such thing as a general rule or a one size fits all. Every single instance has to be analyzed on its own merit and according to its own conditions.
That's why I am very reluctant to join things like BAARF: it's waaaaay to general for my taste.
Now: BAAG, that is a totally different kettle! ;) Received on Thu Apr 07 2011 - 08:31:17 CDT