Watching performance on a DHCP Server
blake at ispn.net
Fri Feb 8 21:27:23 UTC 2008
-------- Original Message --------
Subject: Re: Watching performance on a DHCP Server
From: David W. Hankins <David_Hankins at isc.org>
To: dhcp-users at isc.org
Date: Friday, February 08, 2008 1:38:23 PM
> On Fri, Feb 08, 2008 at 11:32:57AM -0600, Blake Hudson wrote:
>> I'm glad that ISC is continuing development on DHCP, it seems like major
>> gains in performance can be accomplished relatively easily by either
>> combining writes or writing asynchronously. Although, I was a bit
>> surprised that the leases table is not kept (and managed) in memory. And
>> that since dhcpd currently relies so heavily on disk, that it has not
>> taken advantage of any file system or disk performance enhancing
>> features. I look forward to upgrading to 4.1, though I will likely wait
>> until it has been integrated into my distribution for further testing.
> It sounds like you're on the right track, so just a quick note here,
> It is managed in memory - the disk file is just a log that can be
> replayed to rebuild the in-memory database. We think it is imperative
> to be able to reliably rebuild the in-memory database on a reboot for
> various ("no, that's just perfectly normal paranoia, everyone has
> that") reasons.
Seems like it would make sense that this I/O could be non-blocking....
e.g. leases are served out of memory and the log sent to disk. If the
disk cannot keep up, too bad, let it queue.
I would see the value of making this an option on startup, those who are
paranoid can keep an up to date log on disk. Those who have done due
diligence with a battery backed up RAID device and redundant power have
the option to allow the writes to be combined, queued, or otherwise
managed by the filesystem or block devices (which we can assume are
configured to fsync when appropriate).
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the dhcp-users