Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations SkipVought on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Expected Throughput rate on Gigabit Network?

Status
Not open for further replies.

roastbeef73

IS-IT--Management
Jul 14, 2004
5
0
0
US
I'm running BE 9 for a number of Windows 2000 Servers. They all have gigabit NICs and are connected through a Dell PowerConnect gigabit switch. The tape library is a Dell PowerVault 128 T connected to the Media Server (Windows XP) via SCSI. I finally got the system to the point where a 500+ GB backup succeeded. I'm seeing, on average, a 96 MB/min throughtput. Is this a reasonable throughtput rate given the gigabit network and system described above?
 
I wouldn't think so considering I get 10 times the throughput using 100 meg NICs. Check your tape drive properties/configuration - is compression enabled? I get the highest throughput with the settings at the maximum values - block size = 64K, buffer size=1024K, buffer count=32, high water count=32, pass through mode cleared, block mode cleared. Also, is your switch configured correctly?

 
Thanks. I did have compression enabled with the default settings. I changed them to match yours. I'll post if this makes a difference.
 
IF you start changing the compression and block size you could be shooting your self in the foot. If you take those tapes to another server and those changes for block size , etc are different you wont be able to read the tapes.
Couple things you have to try for your network performance.
From the media server browse out to a remote machine Ex \\servername\c%. Copy 1 1gb file to the media server. Copy the file back...do you notice a difference. Update all the network drivers, etc. Do local backup on the media server run as expected?
 
Thanks steveot. I did 6 file copies of a ~1 GB file between the media server (workstation) and three servers each. It was a bit slower copying to the servers than pulling from them which I guess is expected considering the servers each have 3-5 drive RAIDS and the media server is just a single drive workstation. When all is said and done, these 18 file copies averaged 81.3 seconds for a 1,076,101 KB file. Therefore the average transfer speed was 105.8 mbps (1,076 MB X 8 = 8608.808 megabits / 81.3 seconds = 105.8 mbps -- is my formula correct here?). Is this a reasonable transfer rate for gigabit-rated hardware? Any obvious things I should do to improve it?

Anyway, I backed up the same file used above direct from the media server and Backup Exec reported a throughput rate of 96.1 MB/min. Therefore it makes me think that the network is NOT the bottleneck. Where else should I look? Should I modify the compression values?
 
Next thing to try is backup to disk.
Create a backup to disk in the device tab in backup exec.

Run a backup of the servers or one server your getting the 96mb on. Point the job to the backup to disk and not the tape device. If you see a drastic difference then its the tape device. Another thing to try is NTBACKUP and backup to files. See if its really BE or other issues.
 
Yes, test changing the drive configuration to achieve maximum throughput unless you are satisfied with less than optimim speed. I tested multiple settings to achieve maximum throughput on my setup. From the Admin guide (I placed '*' around performance information):

"Block size (per device)

The default is the preferred size of the blocks of data written to new
media in this device.
You can change the block size by selecting another size from the
scroll list, and then clicking OK.

*Some devices (for example, DLT drives) provide better performance
when larger block sizes are used. The preferred block size can range
from 512 bytes to 64 kilobytes.*

Backup Exec does not ensure that the requested block size is in fact
supported by that drive. You should check the drive specifications to
make sure that the block size is supported. If the drive does not
support a block size, it will default to its standard block size.
If the drive does not support block size configuration, this option is
unavailable.

Buffer size (per device)

The default is the preferred amount of data sent to the drive on each
read or write request. The buffer size must be an even multiple of the
block size.
You can change the buffer size by selecting another size from the
scroll list, and then clicking OK.

*Depending on the amount of
memory in your system, increasing this value may improve drive
performance. Each type of drive requires a different buffer size to
achieve maximum throughput.*

Buffer count

The default is the preferred number of buffers allocated for this
device.
You can change the buffer count by selecting another count from the
scroll list, and then clicking OK.

*Depending on the amount of memory in your system, increasing this
value may improve drive performance. Each type of drive requires a
different number of buffers to achieve maximum throughput.*

If you change the buffer count, you may need to adjust the high
water count accordingly.

High water count

The default is the preferred number of buffers to be filled before data
is first sent to the drive, and any time after that if the drive
underruns.
You can change the high water count by selecting another count
from the scroll list, and then clicking OK.
The high water count cannot exceed the buffer count. A value of 0
disables the use of high water logic; that is, each buffer is sent to the
drive as it is filled.

*The default setting provides satisfactory performance in most
instances; in some configurations, throughput performance may be
increased when other values are specified in this field. If you increase
or decrease the buffer count, the high water count should be adjusted
accordingly.*

If a drive has a high water count default of 0, it should
be left at 0.



Jamie
 
I backed up the same 1 GB file from two servers to a "backup to disk" device and the transfers were 927 MB/min and 1001 MB/min which is consistent with the mbps rate I've been seeing for file copies so it looks like the delay is being caused by the tape library or configuration settings. I updated the driver to use the latest from Dell, but I can't find any information on the correct block size settings for the Dell PowerVault 128T. I'm going to update the library's firmware to see if that helps. Any other thoughts?
 
SUCCESS! I've finally got the backup of my 1 GB test file from a networked server to the tape library at a respectable 955 MB/min! I can't wait to see how the actual backups do.

It came down to the SCSI driver. Once I updated that, it flew. Thanks to everyone that provided some suggestions.

Anyway, back to a previous question. Is ~130 mbps a respectable ACTUAL file transfer speed for gigabit networks?
 
Your speed bottleneck may be the tape - is it AIT??

For that much data, you really need AIT-3 or LTO-2.

<signature for rent>
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top