The Key To Performance Metrics
Source: Processor.com – May 7, 2010
By: Holly Dolezalek
Click to download PDF
Unlock Your Enterprise’s Potential By Leveraging The Information You Gather
In any IT environment, there’s plenty to keep track of. Any IT manager can rattle off a baker’s dozen of metrics that he or she could use to measure performance. In fact, the difficulty in succeeding with performance metrics isn’t in coming up with metrics to track. It’s in knowing what to measure, deciding whether those measures have any meaning for the organization, discerning what the data within those metrics means for the health and performance of IT, and figuring out what to do with the data once it’s gathered.
Whatever metrics you’re most concerned with, there are several mistakes to avoid so that you get the most out of measurements without missing their meaning. Metrics are meaningless by themselves; unless you define them appropriately and use the data for action, you might as well not track them at all. Here is a look at what it takes to successfully leverage performance metrics in your data center.
Know What You’re Tracking
What you measure depends on what aspect of IT you’re concerned about. You might be concerned with the performance of your network, for example, and so you might track how quickly the system responds to user interaction. That’s why component metrics—measurements of how the switch, router, server, or other network equipment is functioning—are frequently on an IT manager’s radar.
Application performance is just as pertinent as network performance. For that reason, metrics that track the performance of applications are also on the dashboard. That, in turn, brings metrics of user experience into the mix. Whether it’s a customer-facing application or one used internally, metrics of user experience give context to network performance metrics. “Availability, especially for customers, is key, and it goes beyond simple performance of components,” says Geoff Webb, senior manager of product marketing at NetIQ. “Your system can be running perfectly while some part of your database is thrashing and dying, but if you’re not looking at the right metrics, you can think that it’s running great.”
From the same standpoint, security metrics are another way to keep an eye on system performance, just as system performance metrics can give you insight into your security posture.
“System performance is often the first indicator of an attack or some type of security issue,” Webb says. “When you have a business-critical system that’s performing poorly or a database that’s full or running slowly, the first question should be: Has something changed? If so, what? Was it a planned, managed process? Who did it, and why?”
Monitoring vs. Metrics
Monitoring thresholds isn’t the same as tracking metrics. For example, monitoring a server’s temperature doesn’t tell you anything; it gives you a warning that it’s about to blow, of course, but it tells you nothing about what that server is actually doing. Thresholds are useful for preventing disaster but not for finding efficiencies or spotting developing problems. “In normal business usage, any metric you pick will have its own profile of normal,” explains Webb. “Simple measurements aren’t going to capture that profile.” Webb gives the example of the spike in disk access and process loading in email utilization on Monday morning when many employees come in and start up their email at the same time. If that spike doesn’t occur as expected, it’s likely that a system has failed. But without knowing when that spike usually happens and taking account of it in metrics for email performance, you won’t get the information you need when you need it.
“Taking metrics in isolation as static measures can be very misleading,” Webb says. “So you need to set thresholds, but you also need a way to build an intelligent profile of the way that a system should behave so that you can quickly identify those times when it isn’t performing appropriately.”
Indeed, metrics aren’t useful in isolation. By monitoring your devices, you can get a lot of information about port behavior, bandwidth, and other metrics, but you also need to know how the applications are performing as they pass through those devices. This becomes particularly important when you migrate applications to the cloud, says Steve Shalita, vice president of marketing for NetScout Systems. “Many organizations may have hundreds of applications running, and you have to be able to know about how VoIP, Web traffic, email, video, and other traffics are behaving,” he says. “It’s hard to track meaningful metrics if you don’t know how different services are running and impacting each other.”
Make Your Metrics Count
Don’t track something just because you can. Ron Wince, CEO of Guidon Performance Solutions, explains that his clients often make the mistake of tracking hundreds of metrics. “Sometimes people go ballistic on metrics,” he says. “But it’s better to get down to what’s critical to the business and to prioritize your metrics based on what aligns to the strategy of the business.” For instance, he says, a metric that relates—however remotely—to a company’s market share should take priority over a metric that doesn’t have a direct impact on the business.
In fact, Wince says, there’s nothing wrong with tracking 80 to 100 metrics if that’s what helps keep the business humming. But in that situation, it makes sense to give more weight to some metrics and less to others. “You can weight your metrics in your tracking to give more importance to the metrics that offer an opportunity for improvement,” he says.
Decide on metrics from more than your own perspective. If you’re using the wrong metrics or looking at them without the right context, you can be misled by your own data. “It’s easy to track processor load or uptime and think everything is fine, but with any technology’s performance, what matters is the experience of the user,” Webb says. “You need to look at user behavior and user experience to figure out what metrics will tell you about their experience. That means figuring out what an end user sees, how you can model that, and what metrics will tell you what’s not working right.”
Shalita agrees. “When you know how users are using your systems, you know where the demand is coming from and which application is consuming available bandwidth, because in a lot of cases, performance problems aren’t coming from the network itself,” he says.
All contents © 2010 Processor.com. All Rights Reserved.
Related Links
Guidon Business Process Management Services
Information Technology Industry Solutions
Contact Guidon
Contact us or call us at 1.866.986.4414 or 480.986.4414 (for international callers) for more information regarding how a Guidon solution can help your organization.
RSS
Facebook
LinkedIn
Twitter