How Harmful is Virtual Sprawl?
May 26th, 2016 by Duncan Little
The great thing about having a virtual infrastructure is that creating a new server takes just seconds. Providing you have the resources available you can clone a server for test, development or other purposes pretty much instantly, and over the years this is what many IT teams have done.
When there is a requirement for a new server (even if just temporarily) it is created extremely quickly, but it very rarely gets decommissioned, even after it has served its purpose. This way of working quickly leads to virtual infrastructure sprawl, increasing the amount of servers, and therefore the resources and licences you are consuming.
As we migrate more customers to BrightCloud we notice that nearly all of them, having virtualised several years ago, are suffering from virtual sprawl. However is virtual sprawl on your own infrastructure a problem?
Right-sizing is important
Before migrating a customer into the cloud for our hosting and/or disaster recovery services we often engage in a right-sizing exercise to bring the number of virtual machines (and the associated resources) in line with what they actually need. This right-sizing is good practice when moving your infrastructure into a cloud because it reduces your costs (obviously).
Overprovisioning
But it isn’t just the number of servers you need to be worried about, it’s also the way they are provisioned. Many organisations don’t consider how to group their servers into logical business service groups, instead they provision thick disc that isn’t always utilised and over-provision the CPUs required to run the server and the memory required to run the application.
All of these things can lead to a virtual infrastructure and applications slowing down, increasing costs and a more difficult infrastructure to administrate; particularly if you want to migrate it into a cloud or provision a virtual disaster recovery service.
SPRAWL = increased cost and degraded performance
The BrightCloud methodology
As we have moved so many customers into the cloud, we have developed a methodology to tidy up your virtual infrastructure which reduces your costs and increases performance. This methodology is very straightforward to follow and all IT teams should consider looking at their infrastructure to see how they can reduce costs and gain more performance. Make sure you can answer the following questions:
- How many VMs are duplicated?
This is the easy bit and should only take you ten minutes to work out.
- How many VMs have over-provisioned CPU and memory?
- Are you thick or thin provisioning the disk?
- Why are you over-provisioning?
Is it because of the responsibilities placed on managing thin provisioning?
- Are you utilising the capacity you have provisioned?
This is a very important question when considering storage refreshes and looking at how much capacity you need.
- Have you provisioned that much capacity because you need the spindles to guarantee performance?
Once you have spent the time tidying up your infrastructure you need to put management in place or you will quickly end up back to where you started. However paying attention to what’s happening on the server farm is often a luxury when you are dealing with so many other IT projects and problems. Using tools like vRealize Operations Manager can also be time-consuming and are a little bit too expensive.
This is why we provide a remote monitoring and management service for virtual infrastructures, including other admin tasks such as backup. We install monitoring tools such as vRealize Operations Manager and deliver them to you as a service whilst managing the hypervisor layer of your farm. This brings you tremendous benefits as it not only improves performance, but also saves you money.
