Compaq ProLiant 6000 Compaq DLT Tape Array II: High-Performance Backup of Ente - Page 23

Etwork, Erformance, Imple, Ackup Of A, Emote, Lient, Indows

Get Compaq ProLiant 6000 PDF manuals and user guides View all Compaq ProLiant 6000 manuals
Add to My Manuals
Save this manual to your list of manuals

Page 23 highlights

ECG075.0997 (cont.) DLT Tape Array II ... TEST 8 NETWORK PERFORMANCE: SIMPLE BACKUP OF A REMOTE CLIENT  WINDOWS NT Windows NT 4.0 (server) pulling from Windows NT 4.0 (client) Cheyenne ARCserve 6.0 for Windows NT Single job 2 GB 2:1 RAIT-0 ProLiant 5000, two Pentium Pro 200 processors, 256K cache, 256 MB RAM 80-GB array (two SMART-2/P Array Controllers with twenty 4-GB drives) 35/70 GB DLT drive(s) Wide-Ultra SCSI-3 cards (All Compaq off-the-shelf products) 10TX - Ethernet over CAT-5 cable through a simple hub 100TX - Ethernet over CAT-5 cable through a simple hub All tests used TCP/IP protocol. To document the average times for backing up a remote system over the network at Ethernet speeds of 10 and 100 megabits per second. The results clearly show that simple backups of remote systems are slow, due to the network connection. To provide greater overall bandwidth and minimize the impact on daily network traffic, a dedicated network card in the server is recommended for installations using this type of backup. Consideration should also be given to providing local tape drives to those large systems if better remote performance is desired. The 100TX theoretical maximum is approximately 40 GB/hr, but throughput can be significantly reduced by other factors, such as the protocol used (TCP/IP vs. NETBIOS, for example). If network feed speed is less than the maximum rate at which the target drive can accept and write data on a continuous basis (or stream), the target drive drops out of write mode, and a significant amount of time is lost to shuffling the tape for the next sequential write. The 35/70 GB DLT drive streams at more than 30 GB/hr. A network must run at that speed or greater to prevent the drive from running out of data and stopping. The 10 Mbit Ethernet moves data at less than 3.9 GB/hr, so a single job cannot exceed that, unless an agent is used to precompress the data. With pre-compression, the system might effectively move 4 or 5 GB/hr. During such a backup, the tape drive constantly starts and stops, wasting a large amount of time repositioning the tape for the next write. Although a 100TX Ethernet permits up to 39 GB/hr feed speed, backups in these tests only achieved 19 GB/hr. This is not as fast as local backups because of the processing required to move data over the network. Still, at 19 GB/hr, the network has room for more data, and as seen

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
7
%
(cont.)
DLT Tape Array II
ECG075.0997
T
EST
8
N
ETWORK
P
ERFORMANCE
: S
IMPLE
B
ACKUP OF A
R
EMOTE
C
LIENT
W
INDOWS
NT
Windows NT 4.0 (server) pulling from Windows NT 4.0
(client)
Cheyenne ARCserve 6.0 for Windows NT
Single job
2 GB
2:1
RAIT-0
ProLiant 5000, two Pentium Pro 200 processors, 256K
cache, 256 MB RAM
80-GB array (two SMART-2/P Array Controllers with
twenty 4-GB drives)
35/70 GB DLT drive(s)
Wide-Ultra SCSI-3 cards
(All Compaq off-the-shelf products)
10TX – Ethernet over CAT-5 cable through a simple hub
100TX – Ethernet over CAT-5 cable through a simple hub
All tests used TCP/IP protocol.
To document the average times for backing up a remote system over the network at Ethernet
speeds of 10 and 100 megabits per second.
The results clearly show that simple backups of remote systems are slow, due to the network
connection. To provide greater overall bandwidth and minimize the impact on daily network
traffic, a dedicated network card in the server is recommended for installations using this type
of backup. Consideration should also be given to providing local tape drives to those large
systems if better remote performance is desired. The 100TX theoretical maximum is
approximately 40 GB/hr, but throughput can be significantly reduced by other factors, such as
the protocol used (TCP/IP vs. NETBIOS, for example).
If network feed speed is less than the maximum rate at which the target drive can accept and
write data on a continuous basis (or
stream
), the target drive drops out of write mode, and a
significant amount of time is lost to shuffling the tape for the next sequential write.
The 35/70 GB DLT drive streams at more than 30 GB/hr. A network must run at that speed or
greater to prevent the drive from running out of data and stopping. The 10 Mbit Ethernet moves
data at less than 3.9 GB/hr, so a single job cannot exceed that, unless an agent is used to pre-
compress the data. With pre-compression, the system might effectively move 4 or 5 GB/hr.
During such a backup, the tape drive constantly starts and stops, wasting a large amount of time
repositioning the tape for the next write.
Although a 100TX Ethernet permits up to 39 GB/hr feed speed, backups in these tests only
achieved 19 GB/hr. This is not as fast as local backups because of the processing required to
move data over the network. Still, at 19 GB/hr, the network has room for more data, and as seen
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.