> I've been talking with Trond lately. With the latest version,
> he asked me to try using an nfs mount of a ramdisk on the Solaris
> machine. When I did this, speed was consistent for reads and writes.
> That is, the same 4MB file was written in 0.44 seconds and read in
> 0.52 seconds (I assume the difference here is the speed difference
> in local disks between the one on the Solaris machine and my local
> machine).
This argument half-convinced me that support for rsize>4k is needed
for the Solaris case. I've therefore implemented it in the latest
version (0.7.0) of the NFSv3 patches:
Basically, if you have rsize>4k, each call to the asynchronous read
routine will now attempt to read up to (rsize+PAGE_SIZE-1)/PAGE_SIZE
bytes into the page cache where it will be found by the normal
readahead code. (This affects both the NFSv2 and the NFSv3 code.)
Please could Andrew and others who've been seeing slow reading wrt
Solaris machines test this code out. If it makes no difference, I'll
scrap it...
Cheers,
Trond
PS: The patches against linux-2.2.5 are (as always) available under
http://www.fys.uio.no/~trondmy/src
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/