Taking Control of VMs

By Cameron Sturdevant  |  Posted 2010-07-13 Print this article Print

VMware vSphere 4.1's new features act as a referee when VMs contend for scarce resources.

VMware vSphere 4.1 has a lot more moving parts. Memory compression and the addition of I/O controls for networking and storage, which join those already found in the CPU, are responsible for this increase.

For data center managers, the new capabilities are a boon. The features basically act as a referee when VMs (virtual machines) contend for scarce resources. The additional moving parts also mean more policy to make decisions that will set out the rules used by these newly minted referees.

This increase in VM resource flexibility results in a significant growth in the importance of management tools and skills. To stay ahead of the pack when it comes to squeezing productivity out of virtual infrastructure, management tools will gain importance only when virtualization makes gains in production environments.

This is also true for cloud resources. In fact it may be even more important for cloud-based resources because the decision to spin up or take down cloud-based infrastructure is dependent on automated decision making that must-ultimately will-be governed by the policies people create.

But let me come back to today's reality. With the release of vSphere 4.1, an update to VMware's widely deployed x86 server virtualization platform, data center managers have greater control over how memory, network and storage resources are allocated among VMs.

Memory compression enables IT managers to determine how physical memory will be apportioned between VMs and the physical host to minimize swapping memory pages to disk. Network and storage I/O controls use policy to give some VM workloads priority over others when there is resource contention.

As of today, these memory, network and storage controls are a first step in the direction of increased flexibility. There will be no turning back from this move toward fine-tuning VM resource use.

Data center managers who have been touting virtualization primarily as a consolidation and cost-reduction tactic must now change gears. To maintain the cost savings associated with physical-to-virtual or "virtual first" data center consolidation plans, management tools are required.

Unrelated to VMware specifically, there are several problems that data center managers will confront along the road to a fully managed virtual infrastructure.

First, management generates no revenue. When it comes to cost justification, management is second in difficulty only to security tools. At least, management tools can demonstrate their use on a daily basis by generating utilization reports. Security tools are usually noticed only when they satisfy an audit check box, when they fail and when the renewal bill shows up.

Second, there is a great deal of confusion about the amount of management needed in a virtual environment. A virtualized data center can reduce the importance of individual hardware failures by enabling workloads to migrate seamlessly. However, that underlying hardware fault must still be corrected to maintain the resilience of the virtual environment.

Thus, it's not true that data center managers can just forget about hardware. What has changed is the much lower level of urgency that now needs to be assigned to hardware break/fix operations.

Third, management is a topic that only a mother could love. Management tools force data center managers to operate at a level of arcana that has not been "consumerized." There is very little in the management realm that appears magical. And it's the one area where IT staff must focus only on the negative: What's broken, failing, overloaded or slow.

Very few people on the business side of your organization want to hear about how the sausage is made: They just want to enjoy its savory goodness. But somebody has to be the cook, and whether virtualized or physical, cloudy or sunny, that job falls to IT.

Cameron Sturdevant Cameron Sturdevant is the executive editor of Enterprise Networking Planet. Prior to ENP, Cameron was technical analyst at PCWeek Labs, starting in 1997. Cameron finished up as the eWEEK Labs Technical Director in 2012. Before his extensive labs tenure Cameron paid his IT dues working in technical support and sales engineering at a software publishing firm . Cameron also spent two years with a database development firm, integrating applications with mainframe legacy programs. Cameron's areas of expertise include virtual and physical IT infrastructure, cloud computing, enterprise networking and mobility. In addition to reviews, Cameron has covered monolithic enterprise management systems throughout their lifecycles, providing the eWEEK reader with all-important history and context. Cameron takes special care in cultivating his IT manager contacts, to ensure that his analysis is grounded in real-world concern. Follow Cameron on Twitter at csturdevant, or reach him by email at cameron.sturdevant@quinstreet.com.

Submit a Comment

Loading Comments...
Manage your Newsletters: Login   Register My Newsletters

Rocket Fuel