dan wrote:
> NFS will always be slower(seek times) than a locally attached disk, even
> if all other interfaces are faster.
That's assuming the same filesystem as native on the other end.
NetApp's filesystem will perform better than a typical Linux box and is
likely to be distributed over many drives.
> the locally attached disk will have
> native disk seek times while the NFS will have native + network
> interface + network card + remote bus. this will take an average seek
> time from 9ms and probably double it. for big files this wont be a big
> deal as the difference in time will be negledgible but for small files
> it will be devistating to performance.
For writes, this only matters if you are running in sync mode.
> every IP error, every packet collision, every foot of wire will effect
> backup times.
If your network has _any_ errors it needs to be fixed. Switched
networks have no collisions.
> Though I am still fairly new to BackupPC, I am extensively experienced
> with networking.
If you expect collisions, IP-level errors, or cable-related problems,
you are doing it wrong. These things happen but they indicate something
is broken.
> I suggest you measure transfer times on your local
> system to a ramdisk rsyncing your /etc directory over a about 100 times
> and then do the same thing rsyncing something to a ramdisk on a remote
> server mount via nfs. you will see that on the local system, you can
> sync that directory to ram 100 times in around 1 second or maybe even
> less than that while on the remote ramdisk it takes up to 3 seconds.
> this is taking average disk latency of something like 8ms and reducing
> it to nanoseconds and still resulting in a 3 fold increase. imaging the
> effects of NFS adding full milliseconds to each file.
Agreed, a test running rsync with local and NetApp mounted destinations
would be a good thing.
> as far as putting a monitoring solution on a production server. what is
> the point?
Machines take a certain amount of rack real estate, power, and
administrative work to install/update/repair.
> the monitoring solution likely has a primary goal of
> instructing you when a server or disk goes down but if the machine it
> sits on goes down it cant tell you.
If an operator is watching the cacti or nagios screens, he'll know right
away when the auto-refresh gives a server error.
--
Les Mikesell
[EMAIL PROTECTED]
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
BackupPC-users mailing list
BackupPC-users@lists.sourceforge.net
List: https://lists.sourceforge.net/lists/listinfo/backuppc-users
Wiki: http://backuppc.wiki.sourceforge.net
Project: http://backuppc.sourceforge.net/
