How to Improve ROI on Under Performing Cloud Servers
I just published an article in Data Center Journal (pages 24-27) about the advantages in performance and cost savings of pruning your cloud servers. As I say in the article:
A study by MIT and CalTech suggested that the best way to manage and optimize the performance of multiple cloud virtual machines is to ‘continuously monitor and benchmark’ the performance of the various resources that are provided by a cloud service provider.
This idea has expanded beyond academia and is now being normalized in enterprise data centers, thanks in part to platform performance management tools such as CloudQoS™.
The reason is that cloud servers are not created equal – at least in terms of performance. You can specify the same number of cores and the same amount of RAM on each server in your configuration and get completely different results. Cloud servers are virtualized which means there is such a tight interdependence on CPU, server and networking performance, that one server may have a different component – a hard drive vs. an SSD – and that will make all of the difference. Or the performance difference could be based on resource-hogging workload that is co-resident on one of your servers.
In my article, I discuss how pruning the servers can really boost performance as measured in transactions per second. Here, I would like to provide more details on the ROI involved with a cloud server pruning strategy.
A 600% Return
But for this blog post, let’s look at another case study where the focus is on reducing cost vs. improving performance. This company’s cloud environment consisted of 20 provisioned machines at a cost of $255 per month; for an annual cost of nearly $65,000. A fairly normal cost for this amount of computing power.
We installed CloudQoS on each of those servers and discovered that five of them were under performing. In fact, the workload could be completed on just the other 15 servers.
This leaves the enterprise with a decision to make: They can either “prune” those five servers from their service offering and replace them with other servers that meet their SLA, resulting in a dramatic improvement in their overall performance. Or, since they are running their current workload off of the 15 servers, they can prune the five non-performing servers and not replace them.
Choosing to prune the five under performing servers is what the customer did in our case study. Their per-server charge didn’t change, but their new 15-server configuration meant that their total monthly cloud server spend was $3,825, plus the monthly cost of CloudQoS on each server that totaled about $215. The annual spend is now $48,510; and the ROI on their investment in CloudQoS is 600%.
Cloud services are known as less expensive alternatives to developing internal computing centers, but that shouldn’t mean that enterprises pay too much for their services – especially when CloudQoS makes it easy to know the performance and capability of each server in your configuration. Without the ability to measure the capability of your server, it leaves you limited in how to make confident business decisions on how to optimize your service.
If you are looking to improve the value or performance you are getting from your cloud configuration, then I encourage you to read my Data Center Journal article for more details on our tests.