Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email firstname.lastname@example.org
Re: [Dailydave] The Small Company's Guide to Hard Drive Failure and Linux
From: Derek Vadala (derekcynicism.com)
Date: Thu Nov 18 2004 - 14:34:03 CST
> Make sure you stay away from "hardware RAID" as most of the
> don't even support RAID5 and the performance is really poor 15
> meg/second again 100+ with software raid. Plus you're stuck with a
> vendor with poor support.
Again, what you are calling hardware RAID here is probably firmware RAID.
All the decent vendors can do RAID-5 in hardware. In good cases, this can
be better than software RAID-5 due to caching. On-controller batteries
also help with those pending writes when a power failure strikes. Software
RAID-5 has good read speeds, like any RAID, but writes can be an issues.
Imagine you have a five disk RAID-5 with a 5k stripe size (nevermind that
this is not a valid stripe size). Each time you perform a write, you need
to pre-read the parallel blocks in the stripe you are writing. This causes
a lot of performance overhead for writes, and it's why a lot of people who
can afford to use RAID-10 instead of RAID-5.
> Timing buffered disk reads: 168 MB in 3.01 seconds = 55.79 MB/sec
> rootdis:/home/acz# hdparm -T /dev/md0
> Timing cached reads: 1176 MB in 2.00 seconds = 586.62 MB/sec
> CPU usage is really minimal on this machine (1.8ghz AMD 2500+, 1gig of
> DDR ram),
Again, CPU overhead for RAID hasn't been a factor in a really long time.
This is about I/O, not CPU-- just like nearly every computing bottleneck
these days. The stats you posted are totally useless too. Even if they are
accurate, which I doubt, they are for read performance. You're always
going to get read performance that's near the speed of your real disks.
I'm guessing your disks have a transfer rate of about 55MB/sec. Writes are
what is important. Always. Reads are commodity.
> RAID5 or RAID6 is really the best way to go in terms of security and
> performance. RAID5
> allows 1 drive failure (and will rebuilt its state automatically if you
> have a spare) and RAID6 allows 2 drives to fail at the same time. Using
> other raid modes for anything is pure waste unless you work with big
> temporary files, for which the performance boost of a stripping array
> will come handy; 1 disk failure on a stripping array and you can say
> bye to your data.
Okay. This is one of those things that ends up buried in a mailing list
archive for years, causing confusion among people trying to get good
Most people don't want RAID-6. Most people don't care about surviving a
multi-disk failure on a single array. Most people who need to do that will
probably use a 3-disk RAID-1. That's right, you can create a RAID-1 of as
many member disks as you like. Three disks means you get three copies of
the data, four.... five... etc...
RAID-5 and RAID-6 are not the best way to go in terms of security and
performance. RAID-10 probably is, or perhaps even RAID-1. RAID-5 is great
when you are poor and don't want to waste disk space, but there are
drawbacks. RAID-6 inherits all of that and then some.
Dailydave mailing list