Although the word automation may strike fear in the hearts of some administrators, the term represents a wide spectrum of operational capabilities, from simple, low-level task automation to complex operational processes. But data centers should deploy automation only once they have certainty that automation will do what is intended. This is why testing and validation are important steps in deployment.
Truth be told, it's much easier to implement automation at the infrastructure or server level, because server administrators have full control over the server infrastructure and all the necessary domain expertise. How does a system administrator get started with automation? You can begin by automating simple, repetitive tasks that are easy to document, implement, test, verify and trust before using deploying automation more broadly.
A system administrator could start by setting up an action or series of actions to be launched when a problem arises, such as by stopping and restarting a server or process or by creating a set of diagnostic commands. If a user calls to say that his PC isn't working, for example, a series of tasks to check the problem could be launched (i.e., ping the device; if no response, then ping the server to which it's attached. The series of tasks could be more complex than that as well)..
Although the sequence of actions will be automatically performed once evoked, administrators are unlikely to launch the sequence automatically at first. But they could launch it manually. And once they trust the process, they could launch it automatically in response to a pre-defined error condition.
There are different degrees of automation, and the level depends on the comfort level of administrators with automation and what makes sense for given situations.
This alone can relieve administrators from the many lower-level tasks that consume their time every day. Once these tasks are codified in an automated process, the time savings for administrators adds up quickly. Alternatively, these automated tasks can be delegated to less skilled staff once more skilled staff have codified the steps that must be taken.
This frees up time for skilled administrators that can be redeployed to higher-value activities so administrators can focus on activities that create value (i.e., by resolving issues that can't be automated because they require human intervention and decision making or by delivering new capabilities that are leading edge and valuable to the business).
Remember, automation isn't an all-or-nothing proposition. Use automation selectively, where it makes the most sense and where it gives you the most value at a level of complexity with which you are comfortable. Take advantage of the automation that vendors build into their technologies, some of which are obvious while others are subtle. Some include pre-built best-practice templates to enable rapid deployment, while others deliver automated features. It's an easy alternative to reinventing the automation wheel and figuring it all out on your own.
Automation vendor offerings
The following are a few examples of vendor automation capabilities. As I mentioned above, there are a wide range of automation capabilities: provisioning, configuration, power and heat management, monitoring, performance tuning, capacity, troubleshooting, administrative processes and more. The examples below are just a sample.
Technologies like Opalis and RealOps provide a platform for designing, integrating and implementing operational runbook processes. They allow hierarchical process design that re-use lower level operational processes to create more complex composite processes if needed. With built-in features to assist in building automated processes, runbook automation makes it easier to create and maintain existing operational processes and integrate with existing management tools. Additionally, users can develop new automated processes with pre-built best practices.
If server power usage and cooling requirements are a concern, HP recently announced Insight Power Manager, which measures and regulates HP server power usage. It provides actual power and heat information that can be used to understand power and cooling headroom in a rack. But most interesting is its ability to automatically decrease the processor clock speed during periods of lower processor demand. This saves power and decreases heat output without adversely affecting performance.
When processor demand exceeds 80%, Insight Power Manager automatically resets it to full processor clock speeds to keep pace with the higher demand. Not only does this help administrators manage server rack density by managing power consumption and heat output, you can also derive dollar savings from power conservation, particularly in large data center environments.
Setting server thresholds for alerts is best described as part guesswork, part experience, and part trial and error. Accurately setting thresholds is a balancing act: Set it too high, and you won't know that you have a problem until it's too late; set it too low and your console is filled with false alarms. Netuitive SI dynamically sets thresholds based on statistical analysis and correlation methods that "learn" the normal behavior of the server and detect abnormal behavior. As your operating environment changes, server thresholds are automatically maintained.
These are just a small sample of available server automation capabilities in management products. The breadth and depth of automation-oriented technologies increase every day. Although it's not magic, server automation can feel like it to administrators. And couldn't we all use some of that?
ABOUT THE AUTHOR: Audrey Rasmussen has more than 28 years of IT experience. She served as vice president at Enterprise Management Associates, a systems engineer at IBM, and co-authored the Network World Fusion Network and Systems Management newsletter for several years. Audrey is currently an analyst at Ptak, Noel & Associate.
This was first published in February 2007