mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Maurice Volaski <mvolaski@aecom.yu.edu>
To: support@areca.com.tw
Cc: linux-kernel@vger.kernel.org, erich@areca.com.tw
Subject: Pathetic write performance from Areca PCIe cards
Date: Sat, 25 Nov 2006 15:35:14 -0500	[thread overview]
Message-ID: <a06240400c18e4b03eadf@[129.98.90.227]> (raw)

I have two systems with a serious I/O subsystem based on Areca PCIe 
cards, but the results I am getting from simple write benchmarks are 
extremely slow.

The details are two 64-bit Opteron systems, one with an Areca PCIe 
1210 and the other a PCIe 1220 and both with Seagate SATA II 750 GB 
drives. The motherboard is a Tyan S2891 (Thunder K8SRE). The 120 
system has 1 GB of PC2700 RAM and the 1220 system has 2GB of PC3200 
RAM. The Areca BIOS on both cards is version 1.17a and the firmware 
is 1.41.
I have tried two different kernels, 2.6.17 from Ubuntu, which had the 
Areca driver added by Ubuntu and 2.6.18 from Gentoo with the Areca 
driver added manually from 2.6.19-rc3.

In the initial tests, both computers had a RAID 5/6 configuration, 
but to confirm the result, I setup a single Seagate as a pass-through 
drive and had the same results. The drive was set to SATA II with NCQ 
and the cache was enabled to write-back.

dd if=/dev/zero of=output oflag=sync bs=100M count=1 gives an 
excellent result, around 188 MB/sec.
dd if=/dev/zero of=output oflag=sync bs=200M count=1 gives an 
excellent result, around 167 MB/sec.
dd if=/dev/zero of=output oflag=sync bs=300M count=1 gives an OK 
result, around 117 MB/sec.
dd if=/dev/zero of=output oflag=sync bs=400M count=1 gives a very 
poor result, around 35 MB/sec.
These very low numbers around 30 MB/sec persist as I increase the bs number.

As I continue to run the tests, the bs that gives a poor results goes 
down to about 200 MB. The results are from the system with the 1220 
card. The system with 1210 gives slightly lower numbers overall.

I have also confirmed these low numbers using a benchmark called dm 
from the network RAID package, drbd.

Reading from the drives (based on hdparm -tT testing) gives excellent results.

When I use the drive directly connected to the SATA on the 
motherboard, all the write tests hover around 56 MB/second regardless 
of bs value.

Since both systems are affected, my guess is there is bug in the 
Areca driver or with the cards themselves.
-- 

Maurice Volaski, mvolaski@aecom.yu.edu
Computing Support, Rose F. Kennedy Center
Albert Einstein College of Medicine of Yeshiva University

             reply	other threads:[~2006-11-25 20:35 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-11-25 20:35 Maurice Volaski [this message]
2006-11-27  3:34 ` erich
2006-11-27 13:05   ` Bron Gondwana
2006-11-28 12:19     ` Areca Support

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='a06240400c18e4b03eadf@[129.98.90.227]' \
    --to=mvolaski@aecom.yu.edu \
    --cc=erich@areca.com.tw \
    --cc=linux-kernel@vger.kernel.org \
    --cc=support@areca.com.tw \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®