Does This VM Make My Cloud Look Fat?

Evaluating configured VM sizes in the data center before moving to a public cloud is like trying on new clothes with your spouse. You have some hard questions to answer about size and what you really need.

 

As a married male, the hardest question I often answer is a question asked by my wife on her looks while trying on clothes in the mall. I know, a stereotype, but there is no easy way out of this question. No amount of being nice can cover for making an impulse response when I knew better than to answer the question in the first place. For my spouse, I think the same is true. At some point, we need to be honest and answer right there at the mall instead of after we get home with a bag full of outfits we may never wear.

Admit it, we are not all Abercrombie & Fitch models wearing clothes made just for us. In our data center, our VM is likely in the same situation where they are given resources not ideal for their needs and telling the application owner this information may have the same confrontation experience we might get from our spouse. In the datacenter, it is better to have a tool to tell us and support your answers.

Working in a datacenter and moving to the cloud is a similar task. When we look at moving workloads to the cloud, we need to take a step back and tell the application owners that they may have oversized their VMs, and that oversized configuration is going to cost them month after month until they can shed the excess. Ultimately, being honest and knowing the facts about actual demand and storage needs will make application owners a little happier instead of paying double for resources they do not need.

When you start your cloud assessment process, the first step to success is knowing the actual usage characteristics of your workloads. Long-term assessments will confirm peak usage, but sizing to peak may be adding a little more padding to your VM than is needed. Look at the usage history and analyze your peaks. Did your peak happen on an OS reboot? Did the peak occur during a software update? Did the peak occur during a backup and compression routine during your backup window? All of these events, while important, are not events on which to base your decisions to resize your VMs. 99th percentiles and 95th percentiles start to reveal more characteristics. High 99th and 95th percentiles tell us the VM does in actuality have some heavy usage points. High peaks only confirm our VM has a peak. It is the sustained daily usage of the primary application we need to consider. If a guest OS update takes two or three more minutes to complete, this may not be an issue. If backups still complete within their backup window after rightsizing, do users really care that the process uses 100% of 2 cores instead of 25% of eight cores? At some level, they may willing to take some degradation in performance as long as it is not a degradation in the key application.

So, when I look at a datacenter view of all my VMs, I want to see who is oversized and who is heavily used. Rightsizing ensures I cut the waste, but further analysis also ensures I am not paying for the bloat that is not adding value to my business.

I only wish it was this easy with my spouse. If we would share our opinions more freely, we would know within seconds if the fit was right. I want to find my own ideal fit, but unlike the CloudPhysics Cloud Calculators, I cannot make my clothes fit better in a few clicks. Too bad the real world is not as easy as cloud migration.

Before you commit to the cloud, you may want to try it on for size. When you do, resizing and rightsizing your workloads can save you thousands per year. Take advantage of the CloudPhysics Cost Calculator and Rightsizing tools today before you commit your VMs to the cloud.