linux-raid.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: andy liebman <andyliebman@aol.com>
To: Jim Klimov <klimov@2ka.mipt.ru>
Cc: linux-raid@vger.kernel.org
Subject: Re: Real Time Mirroring of a NAS
Date: Fri, 07 Apr 2006 12:50:18 -0400	[thread overview]
Message-ID: <4436984A.9020803@aol.com> (raw)
In-Reply-To: <8110058661.20060407200418@2ka.mipt.ru>

klimov@2ka.mipt.ru wrote:
> Hello andy,
> 
> al> Can I export NAS B as a SAN or ISCSI target, connect the two machines
> Am I right in the assumption that your NASes are Linux boxes? :)
> 
> Did you take a look at Linux "Network block devices" (nbd/enbd)? They
> might be what you need: you'd get a raw device on one of the servers
> to use in a mirror along with a local device. The NBD page mentioned
> some setups for high-availability services where an active server
> clones itself to a backup server and vice-versa, whichever was active
> most recently.
> 
> I'm not sure about performance though...
> 
> al> with, say, mryinet cards or 10 GbE TOE cards, mount the NAS B volume on
> al> NAS A, and create a RAID-1 mirror of the two volumes? Is this kind of
> al> thing done?
> 
> Are you sure you need 10GbE? My experience with a 10-drive 3Ware 8506
> array in RAID5 shows that reads from it usually fit in 500-700Mbit/s.
> And it's a very busy popular fileserver, so I guess it's close to the
> hardware limits of our array.
> 

Thanks for your reply, and the suggestions of others. I'm going to look 
into both NBD and DRBD.

Actually, I see that my idea to export an iSCSI target from Server B, 
mount it on A, and just create a RAID1 array with the two block devices 
must be very similar to what DRBD is doing, but my guess is that DRBD, 
with it's "heartbeat" signal, is probably more robust at error handling. 
I'd love to hear from somebody who has experience with DRBD.

By the way, I use 3ware 9550SX cards. On a 16 drive RAID-5 SATA array, I 
can get sequential reads that top 600 MBs/sec. That's megabytes, not 
megabits. And write speeds are close to 400 MB/sec with the new faster 
on-board XOR processing. And random reads are at least 200 MB/sec. So, 
10 GbE is a must, really.

Andy

  reply	other threads:[~2006-04-07 16:50 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-04-07 10:47 Real Time Mirroring of a NAS andy liebman
2006-04-07 11:04 ` Joshua Baker-LePain
2006-04-07 16:29   ` Tuomas Leikola
2006-04-07 13:25 ` John Stoffel
2006-04-07 16:04 ` Jim Klimov
2006-04-07 16:50   ` andy liebman [this message]
2006-04-10 18:03     ` Maurice Hilarius
2006-04-19  4:23     ` Tim Moore

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=4436984A.9020803@aol.com \
    --to=andyliebman@aol.com \
    --cc=klimov@2ka.mipt.ru \
    --cc=linux-raid@vger.kernel.org \
    /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;
as well as URLs for NNTP newsgroup(s).