Bill Zumach wrote:
> I'd suggest that the overhead is elsewhere, but not having any idea
> how the 700Meg file is generated or the environment of the test I couldn't
> even begin to guess.
We have an example of copying a large file from the local disk to
AFS and it takes a long time to do the write to AFS.
The file size is 866 Mega Bytes
Copy time is about 22 minutes
Some facts:
The client & AFS servers are running Solaris 2.5.1 (SunOS 5.5.1)
The AFS version is Base configuration afs3.4 5.74
The client cache is 85 Mega Bytes
We use these options for afsd:
"-stat 2000 -dcache 800 -daemons 3 -volumes 70"
I don't think the network, client, or servers are all that busy
doing other work so that is not the limiting factor. The client
and server are on the same Fast Ethernet LAN with only a switch
between them.
-------------------------------------------------------------------
I did an AFS read & local write (copy the file to local disk) using
a different but similar system.
Copy time 20 minutes
I did a read from AFS (I used "wc" to read the file)
AFS read time: 15 minutes
I did a read operation using the local disk file
(I used "wc" to read the file)
Local read time: 11 minutes
Conclusions:
Local disk time is significant for files of this size.
Local disk time is a factor on both the client & the server.
For the 22 minute copy to AFS, there should be a lot of simultaneous
local disk access for the client & server. It seems that local disk
time accounts for about half of the total time.
------------------------------------------------------------------------
| Joe Ramus ESnet, LBNL, Berkeley, CA (510) 486-8683 [EMAIL PROTECTED] |
------------------------------------------------------------------------
