VIRTUALIZATION
Automate monitoring
by Jean-Francois Huard
Companies should understand the performance characteristics of virtualization so they can first put the right management tools and business drivers in place. One area in need of attention is application and performance monitoring.
While existing systems-management tools have traditionally been good at understanding infrastructure availability, they have not been as good about providing visibility into the behavior of resources and application performance. As a result, enterprises might find scaling their virtual and physical data center environments difficult without automation.
Virtualization technology runs much like an application runs on a computer: separate from the operating system, automatically bypassing any hardware and software incompatibility problems. Behavior in virtual environments is constantly changing because of the mobility of the virtual machines (VMs) and the complex interdependencies within the virtual pool of resources.
Managing performance in such an environment requires the ability to constantly adapt to changing behavior patterns. Through automation–and the availability of self-learning and continuously adaptive technology–a level of business intelligence can be achieved that allows for the understanding of system behavior and the anticipation of user interactions.
Since IT supports a large number of increasingly complex and unpredictable user and technology interactions, IT management will need to adapt in the following ways in order to restore predictability and bring performance consistency to virtual environments:
Implement “smart” technology that self-learns. Real-time behavior analysis provides the same benefit of self-learning anomaly detection in the data center. Rather than trying to model constantly changing performance variables, performance management should analyze behavior in real time and correlate infrastructure performance quickly to application performance and vice versa.
Automate threshold management As part of the self-learning capability, thresholds should be adaptive. Performance-management tools for virtual environments should be able to learn and build behavior profiles for servers, VMs and applications, and also adapt thresholds for changing behavior.
Get better visibility into individual VM and system behavior. With so many moving parts in a virtual infrastructure, isolating the cause of an application performance issue can be difficult. Visibility into the health of each VM and the health of the overall system is crucial.
Proactive capacity planning.
Performance-management tools should offer resource allocation and capacity planning before deployment, as well as in production. Self-learning performance-management software has the ability to analyze data and events in real time, and uses these findings to build performance baselines and set thresholds. Most importantly, it is able to constantly incorporate new data so that baseline behavior and performance thresholds are always up to date and mapped to new behavior.
Self-learning performance-management software extends its value in the following ways:
Provides predictable and predictive performance. Because performance management is self-learning, performance is more predictable. In addition, administrators have a baseline of behavior and can identify potential application problems.
Brings about faster problem resolution. These solutions reduce the noise and false alerts generated by disparate management systems. Support staff gets targeted alerts with more context to resolve issues faster.
Optimizes capacity and data center architecture. Accurate performance insight helps IT managers optimize infrastructure consolidation projects in the data center and at branch offices, reducing costs on hardware and data center power and cooling.
Integrates virtual and physical resources. By leveraging existing infrastructure monitoring and VM management tools, the software extends the value of these tools by presenting contextual information and reducing nonessential alerts.
Jean-Francois Huard is chief technology officer for Netuitive, Reston, Va.
For more information
(click here)