IBM ProtecTier Implementation and Best Practices Guide

Chapter 22. Replication and DR deployment with backup applications 399

22.5.4 Tivoli Storage Manager reclamation and housekeeping

When using Tivoli Storage Manager and ProtecTIER CIFS replication, configure your Tivoli

Storage Manager reclamation to avoid the Tivoli Storage Manager reclamation process

colliding with the replication window of ProtecTIER. For maximum efficiency, choose

dedicated windows for ProtecTIER replication and backup server housekeeping afterward.

In the first case, bond all the FSI ports under a single virtual interface (this is also the default

setup). If the user wants to divide the ports among several different virtual interfaces, there

could be a performance issue that using a different subnet per IP address resolves. This is

not a ProtecTIER problem, and is the way the network behaves.

22.5.5 Replication in an FSI environment

You can use the ProtecTIER product to configure replication policies to replicate file system's

directories and all objects contained in these directories recursively to remote ProtecTIER

repositories without any disruption to the operation of the file system as a target for backup. It

is possible to define up to 64 source directories per replication policy, and to define up to

three remote ProtecTIER destinations. The replicated data in the remote destination can be

easily used to restore data in the case of a disaster recovery, or in the case of a disaster

recovery test (without any interruption to the backup and replication procedures).

It is important to allow the ProtecTIER product to supervise all changes made to a directory,

or to a set of directories, that are constantly defined within a replication policy. Therefore, you

should not disable a replication policy, unless this policy is no longer considered relevant. If

there is a scheduled maintenance of the network used for replication, it is possible (though

not mandatory) to

suspend the replication to a specific destination. Suspend allows the

ProtecTIER product to continue supervising all of the changes made; it does not attempt to

send the replication data through the network for the time defined by the suspend operation

(note that the suspend operation is limited in time).

In the case where a policy was disabled for some reason, a new Replication Destination

Directory (RDD) must be defined to re-enable the policy. The ProtecTIER product does not

need to replicate all of the data from scratch if the old RDD was not deleted. It needs to create

only the new structure and metadata within the new RDD. Therefore, you should not delete

the old RDD until a new cycle of replication to the new RDD is over

In the second case, it is alright to have two IP addresses in the ProtecTIER server on the

same subnet, because you have two different hosts, and the issue of using the same subnet

mentioned in the first case is not relevant.

22.6 Replication optimization

This section describes the concepts and procedures related to optimizing replication

performance, including the procedures to improve operations.

22.6.1 General replication performance considerations

A single node ProtecTIER system has two 1 GBps (GbE) dedicated replication Ethernet ports

and delivers a maximum throughput of:

򐂰 190 MBps in physical data-transfer terms

򐂰 1400 MBps replication of nominal data (amount of data before deduplication)