Virtualization is great for streamlining data center operations, but overstuffing servers can easily backfire....
The technology is here to stay, so IT teams must look to strategic capacity planning and make the most of their hardware or end up wrangling with inefficiencies in the future. Virtualization expert David Davis talks about maximizing server space, capacity planning tools and the challenges of capacity planning.
Question: What are some current trends in strategic capacity planning?
David Davis: Companies are trying to make more efficient use of their hardware to get the greatest ROI from their hardware investment by using virtualization. Virtualization allows you to consolidate physical servers as virtual machines (VMs) on fewer hosts. While this is a smart move, it also greatly increases the complexity of strategic capacity planning. No longer do you have a 1:1 mapping of application to server. With server virtualization, you may have a 20:1 (or more) ratio of application to server mapping. Thus, applications can interfere with one another, and you can easily overcommit a physical server by trying to push the capacity utilization to its limits without going over. Capacity planning tools need to understand these relationships in the virtual infrastructure and make recommendations as to how server utilization can be maximized.
Question: What are the top capabilities that strategic capacity planning tools should have?
Davis: Capacity planning tools must do the following:
- Understand both physical and virtual servers. So many tools just use Simple Network Management Protocol (SNMP) to pull back performance data from a host, not really understanding the relationship between physical and virtual hosts.
- Communicate with virtual infrastructure management tools — for example, using the vSphere API. With the introduction of the virtualization hypervisor, capacity tools must not only pull resource utilization statistics from the virtualization management server such as VMware vCenter but also understand how virtual infrastructures work. In vSphere, for example, VMs could have memory overcommitted or disks could be thin-provisioned, allowing the host or data store to run out of capacity.
- Make intelligent recommendations and predictions. By understanding the virtual infrastructure and talking directly to the virtualization management host, capacity planning tools can make more intelligent recommendations such as these:
- indicating how many days before a particular virtual host will run out of memory;
- showing that, by moving a virtual machine from one data store to another, these VMs can achieve better I/O performance; and recommending (and perhaps even completing) resizing of VMs, either increasing vCPU and vRAM or decreasing it to save resources.
Question: What are a few common mistakes in strategic capacity planning?
- Using tools that just do performance and aren’t real planning tools. There is certainly a lot more involved in predicting and planning future capacity utilization than just looking at charts and graphs.
- Not understanding that performance resources are related. For example, when a server runs out of memory, it will start paging to disk, causing slow application response and high disk I/O. When multiple virtual machines (and their applications) are running on the same host (with server virtualization), it can be complex to understand how overutilization of resources by one VM can cause slowness on all VMs.
- Not understanding the applications that run in the data center, their importance to the business and their interdependence.
Question: How will strategic capacity planning tools mature in the next few years?
Davis: Over the next few years, strategic capacity planning tools will mature in several ways.
- Tools will get smarter and more intuitive. They’ll be able to predict the needs of the administrator, such as telling you how long before you run out of a certain resource, or how to balance your resources, or how, if you want to increase one resource (like memory), you could increase performance or capacity.
- Tools will increase their recognition of private and public clouds. Tools will recognize virtual machines and applications as they move from the private to the public cloud (and back).
- Tools will better understand your applications and how they are related. They’ll better recognize application traffic across the virtual network and which resources are talking to one another.
Question: How far can strategic capacity planning tools currently go?
Davis: Capacity planning tools can anticipate the needs (current and future) of the applications running in virtual machines. They can make recommendations on how capacity should be balanced or increased. While very few tools will modify the virtual infrastructure, in the future, some tools will be able to initiate changes to resources in the virtual infrastructure or cloud — for example, provision a new VM in the public cloud to add capacity when needed.
Question: Do you see strategic capacity planning and performance monitoring tools merging?
Davis: Capacity tools need performance data and are really just an extension of performance tools. Yes, capacity planning and performance monitoring tools are becoming one. It’s wasteful to have two separate tools that perform similar and related functions.