[Gllug] Hardware comparisons

Nix nix at esperi.org.uk
Sat Mar 24 12:16:33 UTC 2007


On 23 Mar 2007, Tethys spake thusly:
> "Martin A. Brooks" writes:
>
>>The entire point of hardware RAID is that you hand off this processing 
>>to a dedicated card. _Of course_ a couple of Ghz processor will be 
>>quicker on paper than an on-card controller
>
> Even that's not necessarily true. For example, having a battery backed
> write cache on your RAID controller can improve your disk performance
> by an order of magnitude (depending on workload, naturally).

Of course you generally don't notice this because the kernel caches
writes aggressively anyway.

The major advantage of battery-backed write caches is the assurance of
no silent disk corruption after abrupt power failures in the middle of
RAID-[56] writes. (To date I have seen zero instances of such silent
corruption and have only ever heard of one instance, and that was during
aggressive testing with multiple forced poweroffs. I suspect this is
more a sales point than a serious concern, given the number of *other*
potential problems that can cause silent disk corruption on modern
systems.)

> There are pros and cons to both approaches, though. It's just a case of
> weighing them up and deciding which is best for any given situation.

Unless you're buying for a company and need the bus bandwidth or are
rolling in money, go for software RAID. The increased flexibility alone
is worth it.

-- 
`In the future, company names will be a 32-character hex string.'
  --- Bruce Schneier on the shortage of company names
-------------- next part --------------
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug


More information about the GLLUG mailing list