Welcome!

Linux Authors: Hovhannes Avoyan, Elizabeth White, Carmen Gonzalez, Liz McMillan, Imran Akbar

Related Topics: Cloud Expo, Java, SOA & WOA, Linux, Virtualization, Web 2.0, Big Data Journal, SDN Journal, IoT Expo

Cloud Expo: Article

Predictive Analytics for IT – Filling the Gaps in APM

Predictive analytics solutions for IT can detect, trace and predict performance issues and their root cause

Application Performance Management (APM) grew out of the movement to better align IT with real business concerns. Instead of monitoring a lot of disparate components, such as servers and switches, APM would provide improved visibility into mission-critical application performance and the user experience. Today, APM solutions help IT track end-to-end application response time and troubleshoot coding errors across application components that have an impact on performance.

APM has a rightful place in the arsenal of monitoring tools that IT uses to keep its applications and systems up and running. However, today's APM solutions have some serious gaps and challenges when it comes to providing IT with the entire application performance picture.

Hardware Visibility
Most APM solutions provide minimal information about the hardware and network components underlying application performance, other than showing which components are involved in each part of the transaction. Those that do a better job usually require users to shift to another screen or monitoring system to get more hardware visibility. As with the blind men touching different parts of an elephant, this approach makes it difficult to correlate hardware performance with all the other components driving the application.

The Virtual, Distributed Environment
Most of today's APM solutions were created before virtualization, the cloud, and complex, composite applications took off in the IT environment. With virtual machines migrating back and forth among physical servers at different times of the day or week, and applications dependent on scores of components and cloud services, APM vendors are hard-pressed to provide visibility into the entire scope of a single application.

Predictive Capabilities
As 24 by 7 by 365 uptime becomes increasingly critical to business success, enterprises need to be able to predict and address issues BEFORE they affect the business, rather than after. APM has had mixed success in this area. A recent survey by TRAC Research[1] found that of organizations deploying APM solutions, 60 percent report a success rate of less than half in identifying performance issues before they have an impact on end users.

Enter Predictive Analytics for IT
Filling these APM gaps is how Big Data and predictive analytics for IT can play a significant, highly beneficial role in IT's efforts to maintain application performance. Today, when IT encounters performance issues, it typically has to collect its server, storage, network, and APM folks into a war room to search through mountains of hardware and APM logs, and correlate information manually to isolate the root cause. This resource-intensive process can frequently take hours or even days.

IT has lots of alerts and thresholds to analyze, but those are only as good as the knowledge, experience, and insight of the IT folks who configured them. Just because a server surpassed its CPU utilization threshold doesn't mean that event had anything to do with the root cause of an application issue. Often the real issue is hidden deep in all the delicate interactions among multiple hardware and software components, and may not be reflected in individual thresholds. The same TRAC Research study shows an average of 46.2 hours spent by IT each month in these war rooms searching for root cause. Even more depressing, the root cause is often not found, so IT just reboots everything in the hope that it all works until the same problem rears its ugly head again.

Predictive analytics take over where APM leaves off, harnessing third-generation machine learning and Big Data analysis techniques to efficiently plow through mountains of log data. They discover all the behavior patterns and interrelationships between the IT software and hardware components driving today's mission-critical applications. Over several hours or days, the best solutions baseline the normal behavior of all those components, relationships, and events and use complex algorithms to detect any anomalies that are the early warning signs of developing performance issues. Better yet, because the analytics understand the chain of events involved in the developing anomaly, IT support staff are immediately provided with not only the alert that something is going wrong, but also the behavior of every component involved. This information can shave hours or even days off those war room scenarios. For example, thanks to a predictive analytics for IT solution, a major retailer was able to trace periodic gift card application outages to a misconfigured VLAN. Similarly, a predictive analytics solution reduced - from six hours in the war room to ten minutes - the time it took to diagnose a financial content management performance issue.

Another advantage of predictive analytics solutions is that because they self-learn the normal behavior patterns of underlying components, they drastically reduce the educated guessing that usually goes along with IT staff identifying and setting thresholds against key performance. The inflexibility of these thresholds results in large numbers of false-positive alerts. But with predictive analytics, highly sophisticated algorithms compute the probability of certain behaviors and can therefore generate much more accurate alerts. Some users of predictive analytics solutions have called them the Donald Rumsfelds of IT management tools because they point IT to infrastructure issues they never even knew existed and never looked for. Rumsfeld called these the "unknown unknowns."

However, it is in their ability to be "predictive" that these advanced analytics solutions really shine. By detecting small anomalies early in the game, predictive analytics can alert IT to performance issues and provide enough information to address their root cause before IT or application users even notice them. This can have a dramatic effect on application uptime and performance and a direct impact on user satisfaction and even enterprise revenue. In the case of the document management application, predictive analytics discovered a developing performance issue, and its root cause, the night before it would have affected users placing the application under load on Monday morning.

APM tools have their place in the enterprise, but predictive analytics solutions for IT can kick the effectiveness of those and other IT monitoring tools up a notch by detecting, tracing, and predicting performance issues and their root cause long before any IT war room can.

Resource:

  1. TRAC Research, March 4, 2013: "2013 Application Performance Management Spectrum" report.

More Stories By Rich Collier

Rich Collier is a Principal Solutions Architect with Prelert, a provider of 100% self-learning predictive analytics solutions that augment IT expertise with machine intelligence to dramatically improve IT Operations.

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.