Link Sizing - HP 12000 Design Manual

Table 22 Network Latency
Job Concurrency Required to
Saturate Link
2
4
4
7
7
A new option in firmware versions 3.4.1 and higher is the ability to throttle the number of
replication job that can run on the master node0 of a multi-node configuration.
In multi-node configurations of three nodes or more, this option can be used to prevent
◦
replication jobs running on node0 and thus optimize overall system performance.
In multi-node configurations where you have a separate management LAN from the
◦
replication data LAN, you can prevent replication jobs running on node0:
–
–
–
–
The number of replication jobs on the VLS is set via the "Max number of jobs to run on master
node" option in the Global LAN/WAN Replication Target Configuration. This is a global
setting that applies to the local VLS device (not the remote target device).
When you set the global per job Mbytes/sec limit on the target device, you can calculate the
Mbytes/sec limit on each replication target by multiplying the per job limit by the maximum number
of concurrent jobs set for that replication target. This is only the Mbytes/sec limit but it will only
apply if the replication link bandwidth is higher than this limit.

Link Sizing

This section focuses on understanding how to size the replication link bandwidth required.
The size of the replication link will depend on your usage model for replication.
NOTE: For the volumes of data to replicate with HP VLS systems, HP recommends a dedicated
replication link wherever possible.
Backup and replication can take place in a fixed window (up to 12 hours) preferably when
the WAN link is not heavily used by applications (that is outside business hours). The replication
process has almost the full bandwidth of the WAN link available to it. Within the HP VLS
System configuration for replication you can set a "replication start time and window" during
which time replication is allowed to operate.
The business runs 24x7 and there are no fixed windows in which to do replication. In this
example the replication can be "limited" and the HP VLS devices can be configured to only
use a percentage of the available bandwidth so as not to adversely affect application
performance if applications are sharing the same WAN link.
Network Latency
0ms
50ms
100ms
200ms
500ms
Set the number of replication jobs to zero on both the source device and the target
device.
The separate management LAN must still allow replication management traffic to
pass from node0 on the source device to node0 on the target device in the DR site.
The secondary nodes can then be connected to the replication data LAN that provides
data communications between the source and target VLS.
All secondary nodes must be configured to the data LAN so that replication job
processing is distributed across all secondary nodes.
Aggregate Mbytes/sec
80 MB/sec
80 MB/sec
80 MB/sec
80 MB/sec
52 MB/sec
Design Considerations 1 15