Spent last Friday doing some tests, here is what I have found so far.
Backup Performance Testing
=-=-=-=-=--=--=-=-=-=-=
Baseline
=-=-=-=
Performance Testing with V4.22 client to two different backup servers.
Both were running 7.1.2 on the server side. One server running Suse Pro
9.1 with a 1.1Mbyte a sec tape drive, and a 100MBit lan adapter, and the
other running on Solaris with a 30Mbyte a sec tape drive, and a gigbit
lan adapter.
Results of a full back up of the SYS: volume of a NetWare 6.5 SP4 box
was:
Suse - 1,154Kbyte/sec
Solaris - 746Kbyte/sec
Note: The Suse box ran at the max speed of the tape drive used, and
Solaris ran very poorly. The solaris box is our main backup server and
has performed poorly for netware backups forever. Repeatly packet
traces taken proved there is a packet ack issue, but Legato has never
take taken seriously and never addressed.
Upgrade to v7.2 Client
=-=-=-=-=--=--=-=-=
So same backups of the SYS: volume again:
Suse - 1,163Kbyte/sec
Solaris - 5,412Kbyte/sec
Great news I see a 725% increase in the performance to our solaris box.
Even tho the tape drive is 30Mbyte/sec there is no way this single
server could go that fast as it is hooked up via a 100MBit/sec
connection (10MB). So that it is getting over 50% of the available
bandwidth is pretty good.
StorageNode
=-=-=-=-=-=-=
Previously the v4.22 client did not backup DIRECTLY to a storagenode,
rather it only talked to the main backup server, and that server resent
the data to the storagenode. I retested this, and took a package
capture while the backup was running (16MB buffer worth), the results
are:
StorageNode - 5,191KB/Sec (Oh the storagenode is running on SuSe)
Packet Capture - 99.1% of packets sent to StorageNode
00.9% of packets sent to Backup Server
File Sizes
=-=-=-=-
Perviously the v4.22 client show a big different in performance when
backing up large files, so I wanted to see if I could improve on the
5,412kbyte/sec number. So I have several backup volumes out there that
just have a 1.4GB DVD ISO on it (same one just for testing). I ran the
backup twice, once with a just one volume, and the second with two.
Results:
1 volume with 1.4GB file - 7372KBytes/second
2 volumes with 1.4 GB file - 6963KBytes/second
I consider this test to still be a work in progress, as these boxes
have 100MBit connections, and they have to go thru a router interface to
get to the backup server, so there may not be enough bandwidth to go any
faster and running two at a time may just have caused more delays in the
overhead. I did attempt to hookup multiple 100MBit connections, and the
backup client didn't appear to use the second connection, even tho I had
load balancing turned on. Perhaps with a longer test, or more streams
it would.
=-=-=-=-
Upcomming
1. Deploy client via ZFS to multiple servers
2. Determine minimum number of servers needed to saturate out
30MByte/second drives
3. Restore Test
4. Directed Restore test
5. Cluster Server Backup/Restore - The doc indicate some schedule
restrictions where you can do overlapping backups on the same physical
node. Need to figure out best practice on how to handle this in the
real world...
George A. Raetzke, CNE & LPIC-1
Senior Systems Programmer
Enterprise System Support
Northern Illinois University
DeKalb, IL 60115
graetzke AT niu DOT edu
815-753-8549
To sign off this list, send email to listserv AT listserv.temple DOT edu and
type "signoff networker" in the
body of the email. Please write to networker-request AT listserv.temple DOT edu
if you have any problems
wit this list. You can access the archives at
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
|
