Striking the Right Balance in Server Utilization

Arthur Cole spoke with Mark E. Stumm, senior vice president of product marketing, nlyte Software.

Cole: Recent studies indicate that about 15 percent of the servers in a typical enterprise are drawing power but not performing any meaningful function. Is that a reasonable amount, or should CIOs strive for even less?

Stumm: There are a couple of ways to measure server utilization. There are servers that sit idle waiting for useful work and then there are servers running applications that are no longer needed or used by an organization and are considered “orphaned” or “abandoned.” The 15 percent figure for the number of servers in an organization that are drawing power but not performing any meaningful function is applicable to these orphan servers, and this is a driving factor for justifying the return on investment for data center performance management solutions like nlyte. Discovering and detecting these abandoned servers is the first step in the process of making informed decisions about the best course of action to take. The abandoned application can be moved to a virtual machine, the server decommissioned, and the space, power and cooling capacity recovered to support additional growth. CIOs should strive for an orphaned server percentage in the 1 to 2 percent.

Server utilization for servers running current applications that are idle and waiting for useful work is around 10 to 25 percent on average across all industries, depending on the organization and industry you are measuring. That means that 75 to 90 percent of the time, the server is waiting for work. Now, this is a number that looks really bad on the surface but must be examined in the context of other factors that influence this utilization metric, such as application availability SLAs, performance SLAs, and so on. After considering the performance and availability factors, many under-utilized servers and their applications can be virtualized to improve the overall server utilization for the entire data center, which opens up opportunities to recover capacity by decommissioning or repurposing servers that have been virtualized. The nlyte data center performance management suite ties the virtual environment to the physical hardware on the data center floor, allowing you to visualize your virtual estate and avoid virtual hotspots and sprawl.

Cole: The new version of the DCPM platform mixes asset discovery with workflow management and capacity analysis to streamline data center usage patterns. How do these three components work together to optimize resource efficiency?
Stumm: The nlyte Data Center Performance Management suite was designed to help data center personnel address the problems associated with the “Big Rock” issues inside the data center: utilization, agility and capacity planning/management for things like space, power and cooling. As I mentioned, we would use the nlyte Discover capabilities to pull back utilization information on our server inventory, determine what servers haven’t been accessed in the last 90 days, and what applications those servers are running. Assuming there are some applications we need to move and some servers we can just shut down, we can kick off an appropriate workflow process to execute the application move or server decommission.

These steps in the workflow use different components in the nlyte suite to make the best decisions throughout the process. When the steps are complete and the moves have been recorded in the nlyte database, this information automatically feeds our nlyte Predict module, which uses these trends and historical events, along with planned changes, to accurately predict the future capacity needs for the room, data center or data center estate.

Cole: It also includes something called a "Process Cycle." What is that, and how does it contribute to overall resource management?

Stumm: The Data Center Performance Management (DCPM) process cycle is a systematic set of process steps that we use to address issues in the data center that affect overall resource management. This six-step process (Discover, Visualize, Model, Control, Report, & Predict) promotes continuous improvement and adds a proactive approach to resolving many of the common issues facing today’s modern data center. Each step in the DCPM Process Cycle aligns with one or more capabilities in the nlyte DCPM suite.

The nlyte Discover module collects all IT and facility asset information to accurately record the physical infrastructure of the data center. Once this data has been loaded into nlyte, the Floor Planner and Cabinet Planner modules provide an accurate visualization of the assets on the floor and their physical location in the racks. These two modules also support the ability to model infrastructure changes virtually, accurately assessing the impact of these changes on the available capacity. Once the best course of action has been determined through the modeling step, nlyte Control (Workflow) guides the personnel through the steps of implementing these changes using ITIL and COBIT best practices for data center infrastructure management.

After the changes have been made, nlyte Report provides a variety of canned and custom reports that provide the history of the change and the impact to the facility in real time. The impact of these changes is also rolled into our nlyte Predict module, that uses this historical information and planned changes to accurately predict future capacity needs. After assessing the information in nlyte Predict, the process repeats in order to proactively address resource management in the data center.