Performance testing gluster-integration to optimise sync cycles

[brainfunked]

The default 10 second interval for gluster-integration needs to be validated for it's performance impact. The following setup should be useful for this:

• Tendrl Server node: 4 CPU cores, 12 GB RAM. Dedicate a hard drive to etcd by creating a single partition on the drive with xfs and mount it under /var/lib/etcd.
• A separate prometheus instance to gather metrics from etcd.
• The tendrl server instance needs to be monitored for cpu usage, memory usage, disk i/o and latency and network throughput.
• Gluster setup: 6 node cluster, each with 36 bricks. The bricks need to all be allocated between different types of volumes, total 10.

The testing could be done in a few ways:

1. Create all the bricks and volumes on the cluster and import it into tendrl. Monitor the performance impact during import.
2. Create only a few bricks and volumes spread across all the nodes before import. Post import, create bricks and volumes in batches. Monitor the performance impact both during import and post the creation of these bricks and volumes.

In both the cases, the instance should be kept running and monitored for at least 24 hours.

This testing would allow us to measure the impact on the tendrl server node.

@r0h4n please add your suggestions for profiling the gluster-integration pieces.