Slow backup from remote server to DLT

[Rob Patterson]

We're having a problem w/ the DLT backup of a New HP Exchange 5.5 server. Tape unit is on another server. The Exchange server is the only one having this problem (6.9mb/min). Software is Veritas Backup Exec for NT/Windows 2k (Ver. 8.60).

Have changed out NIC, checked patch cable, switch port, turned off ALL AV software. Doesn't matter if trying to backup store, OS drive, or data drive. Don't recall if system is single or dual proc.

Any thoughts.

Feel free to email me directly at: robp@ntiwa.com

Thanks,

- Rob -

[Andris Mednis]

From our experience with HP-UX and Omniback backup software: if we turn on compressed backups, the data is being compressed by
CPU and then sent to DLT8000 drive - it is slow. Backups wihout compression quickly flow
to DLT, which compresses them quite fast -
it is much better. May be you can check whether
the compression takes place and where.
Also, I have heard that DLTs prefer fast data rates up to 11 MB/s.

[Jaan Kold]

Are you doing inidv mailbox backup? If so..then thats the problem.(the process kills machines and tape drives)
Mailbox backps sound good, and I guess they could be handy, but they are not
supported by Microsoft. See Charles Villa's rant at
http://mail.tekscan.com/nomailboxes.htm for a detailed discussion of why mailbox
backups are bad, as well as some good Exchange Resource links.

And yes, the only supported way to recover individual items is to restore the
entire DS to a standby server and export the items. Doesn't have to be a
monster box, just has to have enough drive space to hold the stores.
(text stolen from Ken Putnam posting)

hope this helps.
jk

[Mark van Hassel]

Hi Rob,

Verify that the NIC and port on the switch is on 100Mb FULL Duplex.

[David Navarro]

Hi, test speed of lan. Many times if full duplex is not correctly set , with 100Mb NIC's speed was very slow. Try with a ftp, ftp gives speed of transfer, and you can check you transfer speed.

[Jay Newman]

Hi Rob.
Not sure of the differences between Veritas Backup Exec and Veritas NetBackup DataCentre, which my company is running.
However, we saw a similar issue on 2 servers.

In one case, it was a network issue (had to force 100 Base T full duplax both on the NIC and the port on the Ethernet switch).
We confirmed the network issue (and later that we had resolved it) by FTP'ing a 5 MB file both to and from the server. (In the initial test, you could send such a file to the server a several KB/S, but the receive from the server was less than 100.)

In the other, the server was starved for resources, and we found much better results by changing the properties of the backup class as follows:
Do not allow multiple data streams.
Limit jobs per class to 1.
In the schedule(s) for this class, set media multiplexing to 1.

Another thought; check your routing tables, make sure the Veritas server and Exchange server. Maybe too many network hops?
Try from each server,
traceroute