There are several ways to increase your server consolidation ratio. Some require new hardware or software, but others you can do on the cheap.
Let’s take a look at five strategies for increasing your server consolidation ratio:
Improving server consolidation with new servers
Replacing an older physical server with a new, higher-capacity server is an easy way to increase your VM consolidation ratio. For example, if you upgrade from a dual-socket server with two cores to a dual-socket server with eight cores, you quadruple your CPU performance.
The new server may also have more RAM than the previous one -- let’s say 32 GB instead of 8 GB. By quadrupling the memory capacity, you may easily quadruple the number of VMs on that host.
Then again, calculating the exact number of VMs you can add to hosts depends on a variety of factors, such as CPU, memory, storage and networking resources. But adding memory and CPU resources will help improve server density numbers.
Also, if you buy a new server with the same socket count as its predecessor, you likely won’t have to pay more for licenses, because a lot of software is sold on a per-socket, not per–core, basis. VMware vSphere, for example, is licensed per socket, and all of the editions up to vSphere Standard support up to six cores.
Improving server consolidation by upgrading memory
In most virtual infrastructures, memory is a constraining resource. If you can’t purchase a new physical server, you can upgrading an existing server’s physical RAM. The price of memory is going down, so buying more RAM for your servers may be more cost-effective than replacing them outright.
You’ll immediately see the server-consolidation effects of a RAM upgrade if you use VMware Distributed Resource Scheduler (DRS). After upgrading a server’s RAM, you can watch DRS load the host with more VMs than it originally had.
Software approaches to improving server consolidation
VMware’s software offers are additional ways to increase your server consolidation ratio, but they may cost more money, depending on your licensing level.
DRS, for one, can make a virtual infrastructure more efficient and help you improve your server consolidation efforts. When a VM is powered on, DRS places it on the host that has the most resources available. DRS also migrates VMs to other hosts when they aren’t receiving the proper resources. By making the most efficient use of your hosts, DRS can help you achieve an optimal server consolidation ratio.
Upgrading to the latest vSphere version can also help. Newer versions of vSphere typically add features that improve performance. VSphere 4.1, for example, added memory compression, which provides faster access to memory when resources are running low. As a result, memory compression allows for a more efficient usage of memory at low levels, so you can place more VMs in a memory-constrained
If you don’t have any money to spend on new hardware or software -- a common scenario in today’s cash-strapped IT departments -- one way to get more out of your existing resources without spending a dime is to right-sizing your VMs.
To right-size a VM, analyze its resource usage and adjust resource allocation accordingly. In many cases, the default RAM and vCPU allocations are more than what the VM actually needs. Reducing the RAM and/or vCPUs assigned to a VM frees up resources for other VMs on the host, creating room to add more VMs. You can right-size VMs manually, but vendor tools such as VKernel vOperations and Embotics V-Commander can make the process quicker and easier.
Setting VM resource limits
Another way to free up host resources for more VMs is to set limits on the resources they use. You can place VMs in a resource pool and limit the resources available to that group, or you can limit individual VMs’ resources.
Resource limiting may reduce the performance of a VM, so take caution with this approach. But there are some low-priority applications, as file and print servers, that can have their resources reduced without end users really noticing. You can also limit the resources available to users who take up more resources than they need to do their jobs (e.g., capping the resources for a knowledge worker who watches videos or downloads large files). By freeing up these resources, you can accommodate more VMs in your existing infrastructure.
This was first published in June 2011