Parallel data transfer using multiple remote mounts

Parallel data transfer by using the multiple remote mounts on the gateway node improves the AFM data transfer performance by sending unique data operations across multiple NFS server mounts.

A cache cluster can be configured for parallel data transfers by using multiple remote mounts when a target site has two or more NFS servers. This feature uses a mapping between the gateway nodes at a cache (primary) site and the multiple NFS serving nodes at a home (secondary) cluster by exporting the target path. When you use NFS as a backed protocol for AFM data transfers, multiple NFS servers can be used at the home cluster. All NFS servers at the home cluster must export the same target path by using the same parameters for a specific fileset.

The AFM parallel mounts feature uses the same mapping infrastructure as the parallel data transfer. The parallel data transfer feature is used to transfer files bigger than the read or write threshold values. However, when the AFM parallel mounts feature is enabled, NFS server nodes can send and receive data in parallel by using the available bandwidth.

In a cache cluster, by using NFS for AFM data transfer, each gateway node can be mapped to a specific NFS server at a home. Creating an export server map can be used to define more than one NFS server and map those NFS servers to specific AFM gateways.

The AFM parallel mounts feature can be enabled or disabled by setting the afmParallelMounts parameter in the mmchconfig or mmchfileset command.

This parameter is available at the global cluster level for the mmchconfig command and at the fileset level for the mmchfileset command.

Do the following steps to configure and set up AFM parallel data transfer by using multiple remote mounts:

Enable the afmParallelMounts parameter at the cluster level.
Set up this tunable at the cluster level to enable the afmParallelMounts parameter for all filesets by using the following command:
```
# mmchconfig afmParallelMounts=yes -i
```
Define a mapping by identifying gateway nodes and NFS servers.
To create a mapping, the user must identify the gateway node at the cache cluster and NFS servers at the home site. This mapping enables mounting remote export from two or more NFS servers at the gateway node and synchronizes unique data by using multiple remote mounts. The user can create an AFM or ADR fileset by using this mapping as the AFMTarget parameter.
Start data synchronization between the home and cache clusters, this synchronization uses multiple remote mounts at gateway nodes.
User can also enable this parameter for a specific AFM fileset. The user can set the afmParallelMounts parameter by issuing the following command:
```
#mmchfileset <FS> -j <AFM/ADR fileset> -p afmParallelMounts=yes
```

Example

The following example shows a mapping for an NFS target. Two cache gateway nodes cachegw_n1 and cachegw_n2 are mapped to two home NFS servers homenfs_n1 and homenfs_n2 (192.168.200.11 and 192.168.200.12), and then SW filesets are created by using this mapping for the parallel data transfer using multiple remote mounts.

Create a mapping between the NSF export server and the NFS server.

#mmafmconfig add mapping1 --export-map homenfs_n1/cachegw_n1,homenfs_n2/cachegw_n2
mmafmconfig: Command successfully completed
mmafmconfig: Propagating the cluster configuration data to all  affected nodes. This is an asynchronous process.

The syntax is as follows:

#mmafmconfig  {add | update} MapName --export-map ExportServerMap

# mmafmconfig show
Map name:             mapping1
Export server map:    192.168.200.12/cachegw_n2.gpfs.net,192.168.200.11/cachegw_n1.gpfs.net

Create filesets by using mapping1 and link the fileset.

#mmcrfileset gpfs1 sw1 --inode-space new –p afmMode=sw,afmTarget=nfs://mapping1/gpfs/gpfs2/swhome 
Fileset sw1 created with id 14 root inode 3670019.

Issue the following command to set parameter for a specific fileset:

#mmchfileset gpfs1 sw1 -p afmParallelMounts=yes
Fileset sw1 changed.

#mmlinkfileset fs1 sw1 -J /gpfs1/sw1 
Fileset sw1 linked at /gpfs1/sw1

By using parallel remote mounts, all the data and metadata operations are uniquely replicated to the target site by queuing them to the unique remote mounts or channels. This method keeps the queue or operation processing completely within the primary gateway node for the fileset and involves multiple NFS servers that are defined in the mapping from the remote cluster. Primary gateway tracks operation that is played to different NFS mounts. This setup supports both write from the cache and read from the remote cluster.

If NFS server failures occur, the AFM detects the failures. The primary gateway node handles the queue and redirects the operation to an appropriate NFS server by maintaining the sync between the cache and home clusters.

Start of change This feature can be combined with the Parallel data transfers feature to obtain better data transfer performance within an AFM cache and an AFM home. Both features use the same AFM gateway node mapping that is defined by using the mmafmconfig command. These features are independent of each other and you can set these features by considering what suits better for a workload. End of change