Cloud Management Platform Cloud Operations Industry Expert vCenter Operations vRealize vRealize Operations vRealize Operations Insight vRealize Suite

David Davis on vCenter Operations – Post #16 – Storage Throughput Analysis

In my last post, I covered how to perform Storage Capacity Analysis using vCenter Operations Manager. In this post, I’ll cover a similar, but different topic – analyzing storage throughput.

Some of the other posts I’ve created around capacity analysis with vCenter Operations Manager are:

vcops-storage-throughput-1

Too many technology people just think about storage capacity when they think about storage. In many cases, it was only when they consolidated their servers with virtualization and put all their virtual machines on a single SAN or NAS (and in many cases, a single SAN LUN) that they realized that the importance of throughput. Unfortunately, too many admins realized the importance of throughput only when they didn’t have enough throughput to keep their critical applications running at acceptable performance. In other words, when they suffered and found that they had plenty of CPU and memory but not enough storage throughput – that was when they began to keep an eye on storage throughput.

Storage Throughput Analysis From the Operations Tab

Like analyzing storage capacity in vC Ops, much of the analysis of storage throughput starts out on the Operations tab. The operations | details overview tab and sub-tab give you a lot of great information about a datastore (or other object that you may be analyzing). From here you have access to a lot of different health badges (see my post on vC Ops badges if you could use some help understanding badges) that give will quickly point you in the right direction if a storage object is having an issue.

For example, in Figure 1 below you can see that the health score of this datastore is 87, caused  mostly by a ~16% I/O utilization rate.  Thus, this array has some files on it that don’t take up a lot of space  (or capacity) but DO utilize a fair amount of throughput.  While a health score of 69 and better is considered “normal”, I would want to watch what I put on this datastore to ensure that I don’t just keep increasing the disk I/O utilization until the array and applications become unusable.

vcops-storage-throughput-2

Figure 2 – vC Ops Operations Details for Storage

If we move over to the Operations | All Metrics tab, you have a long list of metrics to choose from.

vcops-storage-throughput-3Figure 3 – vC Ops Operations All Metrics

It’s here in the All Metrics section that you can individually select which storage-related throughput parameters you want to graph. For example, you can graph things like-

  • Disk command latency
  • Max observed read or write latency
  • Disk workload %
  • and more!

Storage Capacity Analysis From the Planning Tab

Moving from the operations tab to the Planning tab, under Summary, you’ll find all that you need to know to predict and plan for the future of your storage throughput. 

vcops-storage-throughput-4

Figure 4 – vC Ops Planning Summary from the Physical View

Starting under the Summary sub-tab, you’ll see that you can graph both the physical and virtual disk I/O throughput values, as you see in Figure 4 and 5.

vcops-storage-throughput-5

Figure 5 – vC Ops Planning Summary from the Virtual View

Examples of some of the things that you can graph are:

  • Disk I/O read latency or white latency
  • Disk I/O reads per second (or IOPS)
  • Disk I/O reads and writes
  • and more

Storage Capacity Analysis From the VM Perspective

Moving from the planning tab to the Analysis tab, you’ll immediately notice the heatmap view. This view will change based on the disk I/O statistic that you select.

vcops-storage-throughput-6

Figure 6 – vC Ops Planning Storage Analysis

Some of the throughput-related answers that you can find in the analysis tab are:

  • Which VMs have the highest I/O usage and latency?
  • Which datastores currently have the highest usage and latency?
  • Which clusters  have the highest usage and latency?

The idea with the analysis tab is to answer common performance and capacity analysis questions and to show them visually in the heat-map view.

Summary

vCenter Operations Manager is a storage analysis tool for storage throughput. With the types of storage I/O information you are able to view, shown above, you’ll be able to identify storage I/O bottlenecks and bottleneck sources, both for today and for the future.

VMware admins must be prepared with storage I/O data when they need it to solve a problem or to demonstrate to app owners that the problem is NOT storage.

How has vC Ops helped you to solve a storage I/O issue? Post your comments below!

Stay tuned for more great real-world, first-hand knowledge about VMware vCenter Operations Manager!