I/O Workload Distribution - Need Better Visibility

Most storage administrators would like to understand how I/O workloads are distributed over SAN resources such as array directors and ports.  Sometimes, a director or port can be overloaded, causing bottlenecks that can increase end-to-end latencies and impact performance objectives.  It’s also useful to understand the relative loads among sets of directors or ports, to see whether workloads can be more evenly distributed.

ViPR SRM v3.5 has a great new feature that lets you see at a glance how workloads are distributed over array front ends.  This feature is currently available for VMAX and VNX environments.

For example, here is a heat map, available out-of-the box as a standard ViPR SRM report, which gives you a sense of how workloads are distributed across all your VMAX arrays, within a single view.  In this view, all VMAX front end Fibre Channel ports are shown, with each port represented by a colored block.  The relative size of each block gives you an idea of how many I/O requests are being processed by the corresponding port, versus other ports.  If a port is running “hot” (above certain utilization thresholds) it will be shown as orange or red based on defined activity levels.

Here’s a similar view for VNX:

If you want to explore the source of the load on any port, you can do that in a single click. If you click on any port from the heat map, you will be brought to a table showing the physical hosts or ESX clusters that map to that port, with charts showing the I/O performance profile for each of those hosts. 

For an ESX cluster, you can get an idea which VMs are the busiest with a similar view. The following chart shows the relative assignment of compute resources to VMs within an ESX cluster. 

We would very much like to hear from users on how we can improve these views to make them more useful, as well as other areas where you think it would be valuable to employ this view concept.