RE: Database Storage
Date: Wed, 16 Jan 2008 14:39:08 -0500
Just curious. You seem to be a headline reference account for bluearc as of a press release dated December, 2005. What I'm curious about is the vintage and mean time before failure ratings of the spinning rust inside that unit.
Let's presume you're duplex mirroring within the back-end storage. When those units are newish, the likelihood of a two-disk failure of a pair within the mean time to repair the first disk to go bad is something like once in 3000 years. As you get past the mean time before failure ratings of the spinning rust that likelihood rises dramatically, because the expectation of failure per unit time shoots up. Presuming hot swap and silvering or at least hot silvering to available spares, you'll tend to have good experience when most of the devices (spinning rust units) are a lot younger than the MTBF.
I only bring this up to suggest that IF your problem is rooted in an increasing frequency of device failures you'll want to factor in the possibility of replacing the older devices with newer units en masse.
Presumably you have some recent sunk capital cost in replacing individual units.
Of course you have to weigh all these costs against the cost of a new unit. I know next to nothing about bluearc, but if *other than recent disk failures* you've had good luck with them, I'd check if you're just reaching an expected "worn out" problem. Usually companies will be quite reasonable about special deals to update technology, especially reference accounts.
On the other hand, if your problems bear no semblance to "worn out" issues, stuff being unreliable is not a good thing for databases. It would surprise me if any vendor in this situation would not dispatch a crack team to analyze your problem.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Mir M. Mirhashimali
Sent: Wednesday, January 16, 2008 2:08 PM To: Jeremy Schneider
Subject: Re: Database Storage
No The database is fine. The back-end storage we are currently using has been so unstable that we had to shutdown the database several times this month. and couple times it just rebooted itself corrupting the database.
We have a Titan storage solution from a company called Bluearc (www.bluearc.com). so i have been tasked to find out what other people are using as a storage solution and how stable it is. we have redundant servers to handle high availability (not RAC though) manual switchover in case primary server fails.
Jeremy Schneider wrote:
> Just curious, but what do you mean by "performing badly in terms of
> availability (but not throughput)"? Do you mean that the database is
> very fast but crashes a lot? You mentioned an active-passive cluster
> - is it failing over? If so, exactly how often?