Optimizing Physical Infrastructure to Get More from Virtualization and the Cloud

Patrick Donovan is a Senior Research Analyst with Schneider Electric’s Data Center Science Center. He has over 16 years of experience developing and supporting critical power and cooling systems for Schneider Electric’s IT Business unit including several award-winning power protection, efficiency and availability solutions.

pdonovan_tn

Our increasingly digital lives cause data to grow exponentially, while increasing the need for more compute resources. Virtualizing workloads can harness more computing output from IT hardware. But while the benefits are well known and being taken advantage of, their effects on data center physical infrastructure are less understood and often overlooked.

Optimizing and properly managing the data center’s physical infrastructure systems in light of the four potential impacts brought on by a highly virtualized, cloud-based environment is a critical consideration for IT and data center managers. Proper management of these impacts not only helps operations remain unaffected, but can significantly extend the efficiency gains achieved.

Potential Impacts of Virtualization

One such impact is the likely rise in localized areas of high power density caused by virtualized machines (VMs), which require increased processor and memory resources and generate higher power consumption than non-virtualized machines. These virtualized, high density servers are often grouped together creating “hot spots” in the server room that lead to cooling problems.

To combat hot spots, data center managers can choose from a wide array of solutions. The most common method is to spread high density equipment throughout the server room. However, for managers that oversee a smaller data center or operate on a tight budget, this may create space problems. One way to better leverage a smaller space is to cluster all high density servers into one pod, which would consolidate the VMs into one rack or row of racks. Using dedicated cooling air distribution or containment also allows for better space utilization, increases efficiency, and helps obtain maximum density per rack. Additionally, when the air that streams into these pods is well contained, the high density pod may likely add cooling capacity to the rest of the data center. However, a second lesser-known change brought on by virtualization and the migration of applications to the cloud is the likelihood that power usage effectiveness (PUE) will deteriorate.

Optimizing Power and Cooling

If the power and cooling infrastructure remains as it was before virtualization was implemented, then PUE will worsen after the physical consolidation of servers and storage has taken place. Inherent in unused power and cooling capacity is what is known as “fixed losses,” i.e. power that is consumed by the power and cooling systems regardless of the IT load. The more power and cooling capacity that exists, the more fixed losses will be realized. As the IT load shrinks from consolidation these fixed losses become a higher proportion of the total data center energy use, and PUE will diminish.

To improve post-virtualization PUE and lower costs, the data center’s infrastructure efficiency curve must also be improved by optimizing power and cooling systems to better align capacity with the new, lower load. Optimized power and cooling infrastructure should incorporate the following design elements to minimize fixed losses and maximize the electrical efficiency of the virtualization project:

Although likely to have the biggest impact, lowering power and cooling capacity may be the most difficult to implement for an existing data center. Reducing these systems’ capacity may not be feasible in certain situations such as when the design is not modular, if the infrastructure systems are new, or when the dependence of other systems or operations may preclude the time required and ability to reduce its capacity. In these cases the costs of rightsizing the power and cooling infrastructure may outweigh the benefits of closely aligning supply with demand. For an existing data center, more feasible options might include:

For a data center that is currently in design, right-sizing the entire power and cooling plant to the load makes more sense. Doing so at this phase means lower initial upfront capital costs and much better energy efficiency once the data center is operational.

In a data center, electrical loads on data center physical infrastructure (DCPI) can vary in time and in place, so are able to move from one location to another. However, a highly virtualized, cloud-based data center has larger load variations compared to a non-virtualized one, which could cause capacity issues.

To combat the sudden movement of IT loads, managers are increasingly automating the creation of VMs. Automation helps virtualized data centers become more fault-tolerant, enabling one machine to pick up where another has crashed, with limited latency experienced by the user. Despite this development, the rapid movement of VMs can also expose loads to potential power and cooling problems and put the load at risk. Implementing data center infrastructure management (DCIM) software can help alleviate these issues.

DCIM software monitors and reports the health and capacity status of a data center’s power and cooling systems, IT equipment and DCPI. Some DCIM systems can report this information to a VM manager. Without this information, it is nearly impossible for VM managers to ensure VMs are being created in or moved to a host with healthy power and cooling resources. An added level of performance protection comes from automating both the monitoring of DCIM information and the implementation of suggested actions - greatly reducing the risk of downtime in the event of a sudden loss of power redundancy or cooling problems. To automate both steps, DCIM software can be integrated with a VM manager, which allows VMs to be safely and seamlessly moved to safer areas of the room should a power outage occur without any staff intervention. This type of integrated software becomes more important as power and cooling capacities are scaled to match newly virtualized data centers. Moreover, right-sizing infrastructure such as power and cooling may also lead to a reduced need for redundancy in the data center’s physical infrastructure.

Right-sizing the Physical Infrastructure

The reduced need for power and cooling capacity is a widely known benefit of virtualizing and moving to the cloud. A similar, but more overlooked, benefit of this consolidation is the possible reduced need for physical redundancy. Relying more on the fault-tolerance of well-managed VMs and less on higher levels of infrastructure redundancy could simplify design, lower capital costs, and save space for other needed systems or for future IT growth. Before making these types of decisions, IT and facilities management should consider the possible impacts to business continuity if the DCPI or component being considered fails or becomes unavailable.

Also, IT management systems and policies should be reviewed and monitored to ensure they are capable of providing the level of service and fault tolerance that permits less redundancy in the DCPI. Right-sizing in this way can further reduce energy consumption, capital costs, and fixed losses all while improving the data center PUE.

In summary, virtualizing a data center’s IT resources can have certain consequences related to the physical infrastructure. If these impacts and consequences are ignored, the broad benefits of virtualization and cloud computing can be limited or compromised, and in some cases, severely so. Implementing the tips and approaches described in this article will keep a highly virtualized data center running with greater reliability, efficiency, and expanded flexibility to meet highly dynamic compute power demand.

Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.