Re: Informix and RAID

The argument of loosing a second disk is not a very good one. With the current MTTF of disks these days the chances of loosing 2 disks in a array in the same day are very slim. I've heard histories that people didn't notice that the first disk went bad for weeks until the second one went under . I've heard stories of people complaining of raid-5 performance was bad only to find out later that the write-back cache was never turned on in the first place. Blame the people not the technology. RAID-5 might be a good solution in certain applications. Evaluate its pros and cons and how it compares against your other disk options. Based on that take your decision.

Db Solution wrote: > The argument of loosing a second disk is not a very good one. With the > current MTTF of disks these days the chances of loosing 2 disks in a > array in the same day are very slim. It is to laugh! I was going to fill a few lines with HAHA's but that would be rude and non-productive. Let me just say that you are WRONG! It happens. It happened to Obnoxio, it happened to me and I know of at least three other sites where it's happened. The last one to tell me about it messaged me last month. Yes MTTF is HUGE these days, but that's Mean Time Between Failure not Minimum Time Between Failure. Besides have you EVER heard of an SA who swapped out a drive because it's MTTF had expired? Beyond that there are other reasons besides wear and tear for a disk failing including manufacturing defects. That's the one that bit us. We had 200 drives from a single manufacturer's lot installed out of an order of 500 when the vendor called to say the entire batch was bad and that the spindles would seize when they hit a certain installed life (I think it was ~2000 hours, about 3 months) and we'd started installing those drives the previous quarter. After 2 weeks during which our techs and the system vendor's techs worked all night swapping out and rebuilding RAID5 drives we managed to replace all 200 drives from the batch. However, almost 100 drives failed during those two weeks and during the couple of weeks just prior to that. And yes, on one RAID5 structure we did have a multiple failure and only the backups saved the filesystem. Fortunately there was no data loss, but if that structure has been one of the ones we used for our Informix servers (we were running all databases without logging at the time much to my chagrin) we could have lost the whole sheebang. It was immediately after that the management finally took two pieces of advice I'd been harping on for 5 years: They let me add logging to all our databases (after making me fix all the software to deal with the new locking and transaction issues) and the firm no longer uses RAID5. Today we only use RAID10. As a bonus we avoided hardware upgrades for two additional years saving millions of dollars and were able to go to the next generation hardware in one step instead of two. RAID10 paid for itself immediately and then over and over. To have survived this with RAID5 was a miracle. Using RAID5 it would have been a cake walk because we would never build both sides of any mirror with drives from the same lot. To do the same for a RAID5 array we'd need drives from at least 6 different lots (n drives plus the hot backup) in every array and we have hundreds of arrays. Art S. Kagel > I've heard histories that people didn't notice that the first disk went > bad for weeks until the second one went under . I've heard stories of > people complaining of raid-5 performance was bad only to find out later > that the write-back cache was never turned on in the first place. Blame > the people not the technology. > > RAID-5 might be a good solution in certain applications. Evaluate its > pros and cons and how it compares against your other disk options. > Based on that take your decision. >