[mythtv-users] OT Hardware RAID 5 Recommendations
marshalj at spots.ab.ca
Tue Dec 12 20:41:02 UTC 2006
Actually, if you run the benchmark and then run 'sync' afterward, the sync
shouldn't return til the write buffers are all flushed. Unless there are
write buffers on the hw RAID5 controller; then all bets are off!
In general, professional disk benchmarkers use files that are 20+ times
larger than RAM to obviate caching effects. I learned that from the nice
people at 3ware when I was evaluating some of their stuff.
> Interesting, How much ram does the box have? The reason that I ask is
> that you have to make sure that you dd a file that is large enough to
> cause the machine to actually start writing to the disk instead of just
> caching the writes to ram. I've run into the situation where a machine
> with a large amount of ram, 4GB to be exact, wouldn't return consistent
> results till I used larger file sizes that caused the write caches to be
> Paul Hamm wrote:
>> Interesting ran this on my machine. See the bottom for PC specs. The
>> system is running 2 instances of FAH. Not a very consistent test.
>> Using a 6 Disk raid 5 array.
> mythtv-users mailing list
> mythtv-users at mythtv.org
| Jason Marshall, marshalj at spots.ab.ca. Spots InterConnect, Inc. Calgary, AB |
More information about the mythtv-users