Running servers at a quarter or less of their capacity is like running a pizza delivery service with 20 ton trucks. It just doesn’t make sense.

Ask anyone in IT what he or she believes is required to run a particular workload and they will advise you to take the latest and greatest model available. They will do this for a variety of reasons, but mainly because any performance issue creates an overwhelming workload, angry managers and stress by the bucket loads.

Around 10 years ago, people discovered that many servers were only utilized at 15 to 25 percent capacity. The industry responded by implementing virtualization software. The promise was by using this technology we could save on hardware and still run the same application simply by utilizing the existing hardware power.

Seeing the whole picture

As with most IT trends, we found that this was not simply a hardware issue – it was also a people one. We find ourselves left with distributed servers that are not utilized in the best way possible.

Now we are seeing that even many of the virtualized systems have too much capacity for what they need to do. A recent survey showed that 80 percent of virtual machines are oversized. This means saying goodbye to hardware savings.

What makes things even more difficult is that we also found that there is a scary unbalance between memory utilization and CPU utilization. This again demonstrates that simply adding memory would solve more than adding more power would.

Find bottlenecks today to optimize performance tomorrow

This is not to say that I am disregarding how difficult scaling hardware or finding the right fit for an application is. It is very difficult. In fact, it’s almost not doable unless you are using the right software.

With this in mind, a good place to start is to be aware of how things are run today, to find bottlenecks and tuning opportunities. Combined with intelligent capacity management, this will help you to decide what the optimal configuration is for a given application. This will result in you having a real solution for an oh-so-real problem.

So, let’s look to the future. Imagine that you have both application performance management and capacity management implemented and running. Finally, you too can give an answer to questions such as, “Our new app will generate the equivalent of 1,000 users of workload, do we need extra hardware?” or, “We are upgrading three old servers with new ones running eight-way super-duper processors and 25GB of memory, how many existing servers can we migrate to those and how will this impact performance?”

Never again will you have to move your finger to the bottom of the pricelist to make sure that you will not run into performance or capacity issues. And never again will your CIO come to you and ask you: “Why do we have 2,500 servers when none of them are utilized to more than 20 percent. I feel like I am running a pizza delivery service with 20 ton trucks!”

What a bright picture the future is set to be.

Image credit: Anders Johansson

The post Do more with less hardware appeared first on Highlight.