:reposting to the hackers
:;-)
:----- Original Message -----
:From: Robert Gordon <[EMAIL PROTECTED]>
:To: <[EMAIL PROTECTED]>
:Sent: Thursday, December 09, 1999 1:56 PM
:Subject: Clustered Read/writes and NFS..
:
:
:> Hello,
:>
:> I'm attempting to understand if the NFS implementation takes
:> advantage of clustered read/writes. So far I see that if NFS needs
:> a buf that (via getnewbuf()) a call could be made to vfs_bio_awrite()
:> which could cause a clustered write to free up some buffers... but I
:> don't see that NFS takes advantage of a clustered read/write....
:>
:> Thanks,
:> Robert........................ [EMAIL PROTECTED]
Unfortunately there is no advantage to clustering NFS writes, because
the NFS write packets are limited to the NFS write size which is usually
8K, so you still need to do separate RPCs.
But there is an advantage to clustering NFS commits (essentially the
write-verification part of an NFS write), and FreeBSD *does* do that.
In fact, Alfred is working on a patch that will make the commit clustering
even better.
Under NFSv3 the NFS server may cluster NFS write packets it receives prior
to writing the data to the physical media, and FreeBSD does do this.
This is because NFSv3 writes are allowed to be asynchronous. NFSv2 writes
are required to always be synchronous which prevents the server from
clustering them when transfering the data to physical media.
-Matt
Matthew Dillon
<[EMAIL PROTECTED]>
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
