Quoting Larry Kingery (larry.kingery AT veritas DOT com):
> First, let's assume that the tape drives, SCSI, etc are not the
> bottleneck, and that your Gb card and switch port settings are all
> configured correctly.
They are. :)
>
> What is the network layout? If there's a 100Mb connection between two
> routers that all of your clients have to cross, that Gb isn't going to
> do anything for you.
There is not. The network is not the issue. The router is bored with the
amount of traffic I'm sending it.
--------
machine1-- NFS --- 100mb --> |router|
machine2-- NFS --- 100mb --> | |----- 1000mb ---> Media Server
--------
> BTW, what you're trying to do is mutlistreaming, multiplexing is
> when you write multiple streams to the same drive at the same time.
I am trying to send multiple streams to the same drive, so I am, in
fact, multi-plexing. What I'm doing is treating each NFS server as a
NEW_STREAM.
NFS
machine1:/data ------------> MediaServer:/opt/machine1/data
machine2:/data ------------> MediaServer:/opt/machine2/data
The file list for the backup policy looks like this:
NEW_STREAM
/opt/machine1/data
NEW_STREAM
/opt/machine2/data
Thus, even though it looks like I'm trying to create two streams from a
single location (i.e. /opt), in effect, those are actually pointing to
where I've mounted the two NFS servers.
It turns out this does work after all. I neglected to increase the max
jobs per client via the gui, so it was starting one stream and queing
the other.
Thanks for the pointer to the O'Reilly book,
-odeegan
|