Re: Oracle on Digital Unix RAID

Steve Bourgeois wrote:
> *sheesh*
>
> I hope you would be a little nicer to the poster if
> he stopped by your office for your opinion...

I didn't know I was being cruel. But he'd get the exact same response in person.

> RAID 5 typically does not equal the performance of RAID 0+1 because of
> the overhead required to calculate and write parity data. Write-intensive
> data files, redo logs, etc will perform better under RAID 0+1.

Da. Key="typically". Fortunately there are good implementations of RAID whereby response to any configuration of the disks is roughly on par. Given an equality of two monster disk units, one config'd as 5 and one as any noparity  config, and assuming no intelligence in the disk unit, yes, the 5 will be slightly slower on long writes. There are simply too many variables to make sweeping judgement statements like "RAID 5 sucks". My point being, in castigating aforementioned net.newbie, that if you give your opinion you'd better back it up with facts.

> Agreed. RAID 5 is a good compromise in regards to providing redundancy for a
> reasonable cost, but it is not always the answer...

I don't recall saying it was always the answer. I said it was unbeatable for a set of values I stacked in it's favor.

Select count(users) from world <= select count(opinions) from world

No, it's not always the answer, but it does work, and rather well.

bmartin Received on Wed Sep 04 1996 - 00:00:00 CEST