Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 30 Mar 2003 12:01:17 -0500
From:      Chuck Swiger <cswiger@mac.com>
To:        freebsd-stable@freebsd.org
Cc:        Lukas Ertl <l.ertl@univie.ac.at>
Subject:   Re: vinum performance
Message-ID:  <3E8722DD.5050703@mac.com>
In-Reply-To: <20030330175605.E23911@leelou.in.tern>
References:  <20030330125138.K23911@leelou.in.tern> <3E870CC7.5000204@mac.com> <20030330175605.E23911@leelou.in.tern>

next in thread | previous in thread | raw e-mail | index | archive | help
Lukas Ertl wrote:
[ ... ]
>> There are three goals or priorities to choose from when configuring
>> RAID: performance, reliability, and cost.  What are yours?
> 
> I just wanted to test the performance of these drives and of vinum; I had
> no goals to reach.

Testing something to see what happens is a goal in and of itself.

Nevertheless, my question wasn't idle: you probably will find that if 
you choose a goal like, "I want to set up a RAID volume that has really 
good performance", that you learn more from your testing.

>> Also, what tasks you intend to use the RAID filesystem for are critical
>> to consider, even if the answer is simply "undifferentiated
>> general-purpose storage".  In particular, RAID-5 write performance is
>> going to be slow, even with RAID hardware support which offloads the
>> parity calculations from the system CPU(s).  RAID-5 is best suited for
>> read-mostly or read-only volumes, where you value cost more than
>> performance.
> 
> Ok. But I still don't understand why RAID 5 write performance is _so_ bad.

RAID-5 trades performance for both cost and for reliability.

Actually, most forms of RAID trade performance OR cost for reliability, 
but RAID-5 does both and thus is generally slower than a single drive. 
Especially when you're got lots of multithreaded small writes; that is 
to say, doing an I/O benchmark, which is designed to saturate the I/O 
system deliberately, and tends to have an even balance of reads to 
writes, is pretty much a worst-case usage scenario for RAID-5.

That doesn't mean that RAID-5 isn't useful, and it can perform okay 
under light to moderate I/O loads, but RAID-5 degrades badly under high 
write loads.

> The CPU is not the bottle neck, it's rather bored. And I don't understand
> why RAID 0 doesn't give a big boost at all. Is the ahc driver known to be
> slow?

FreeBSD supports Adaptec hardware very well, in general.
Or is that vice versa?

>> Um, that is a dual-channel card, and you're splitting drives onto both
>> channels, right?
> 
> Yes, it is dual channel, but the disks I'm testing are all connected to
> the same channel. Bad layout?

Highly non-optimal?, depending on how you wish to look at it.  :-)

-- 
-Chuck

PS: You should also consider what happens if a drive fails; what happens 
to the performance then?  If you've got hot-swappable hardware, try 
yanking a disk or two...



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3E8722DD.5050703>