What exactly is performance?

To answer her question, we need to define performance. It's a software-defined era, so performance needs to be redefined as the architecture has changed.

In Chapter 3, SDDC Management, we leaned that there are two layers:

• The VM layer: This is where her complaint lies
• The Infrastructure layer: This is where you are looking for the solution, as you are not familiar with her application and do not have access to the VM

You see the problem?

Yup, there is a major disconnect. You are trying to help in an area that you neither have access to nor knowledge of. You have no expertise in the application, and you don't have full admin rights on the Guest OS.

Is there hope?

There is, if you are willing to let go your status of a superhero. Ponder the new nature of your business. In SDDC, you are an IaaS service provider. In HDDC, you are a system builder.

At the core, an IaaS provides these four basic resources to a VM:

• CPU
• RAM
• Network
• Storage

You need to ensure that you deliver the amount of resources that the VM asks whenever it needs them.

Your IaaS is performing fast if it delivers the amount that the VM asks for within your performance SLA. If you promise a 20-millisecond disk latency, your IaaS is doing well if you return all the I/O issued by the VM within 20 milliseconds. Whether 20 milliseconds is good enough for the custom application is not within your scope. If you promise a maximum of one dropped packet in any given 5-minute interval because you are catering for vMotion, then your IaaS is performing if it is only dropping one packet during the period where the VM was not performing. If your IaaS dropped two packets in any given 5-minute period, you have failed your performance SLA.

You are in charge of your IaaS performance, not her VM performance. Your IaaS can perform fast, even exceeding the SLA you promise, but that does not guarantee that her VM meets her performance criteria. Generally speaking, application-level performance SLA is business centric (for example, 8 seconds per financial transaction), so it is beyond the control of your IaaS.

Let's use a table to summarize what exactly IaaS performance is:

Once you determine that your IaaS is serving her VM well, you are done with your scope. You may decide to come back to your hero status and help her troubleshoot that in-house custom application that you're not familiar with. That's a value-added service from you. Your IaaS stops at Infrastructure.