Managing application performance is complex and ever changing. To date, IT has been forced to react to crises as they occur and is measured by how long it takes to fix a problem – MTTR (mean time to repair) – not by how well they can predict or prevent issues before the user even knows they exist.
That world is changing.
Complexities of cloud migration, mobility, BYOD (“Bring your own device” to work), and other initiatives are forcing IT to move beyond merely identifying and repairing issues. IT must now actually predict the time needed to anticipate problems and maintain optimal performance of mission-critical applications. In fact, there are early indicators of our ability to even prevent application performance issues before they occur.
Navigating the minefield of MTT Acronyms – Mean-Time to Identify, Mean-Time to Repair and the latest, Mean-Time to Prevent – can be a real headache. Each has its value and proper place in the technology operations landscape. The challenge lies in how to best leverage breakthrough technologies. IT is being driven to actually save an end user's experience, before a problem has a chance to occur.
Predictability has become paramount and analytics have moved beyond basic time-based correlations. The next generation of analytics tracks behavior patterns across the multi-dimensional nature of today’s application usage. And, those analytics must be automated, not dependent on manually set IT triggers. It’s just too complicated now.
Not long ago, car enthusiasts would spend time under the hood juicing their “hot rod's” performance. Now, with the incorporation of complex chipsets throughout the vehicle operations, those days of tinkering are gone. The same is now definitively true with most of today’s enterprise-grade application stacks.
There is a whole new breed of automated, real-time predictive analytic technologies changing the APM game, especially in cloud-based and hybrid environments. So, let’s take a peek “under the hood” of the wave of APM breakthroughs that allow IT to focus on innovation and business-building activities, not chasing fires, how they can stop fixing and start predicting and then preventing operational distractions, and how users can truly trust the predictability of an application’s performance.
There are clearly a number of hyped trends swirling about these days – “Migrating to the Cloud,” “Consumerization of IT,” “Hybrid IT Environments,” “BYOD” and more. But, with each those game-changing innovations, new challenges sneak into the system. Challenges that pose a threat to end-user experience can be the most difficult to diagnose and to address quickly and completely.
This is where the concept of predictability becomes so significant. Many have used traditional time series data analytics to diagnose and set safety warnings or alert thresholds in place – but they simply don’t take into consideration the complexities and dependencies introduced by today’s evolving networked, distributed applications.
One interesting approach making a big impact leverages load-behavior learning across all points in a networked, distributed application to truly understand how slow or fast each of the gears should move to deliver the end user’s service level targets. It’s not just behavior learning (BL), which has a single dimension – tracking behavior of one metric with time - but takes into consideration the multi-dimensional nature of application usage. Multiple users executing multiple transactions – called usage patterns – that can change every minute.
Application Behavior Learning (ABL) analytics capture and analyze the real-time performance data of each tier of an application infrastructure in relation to the load generated by real end users. ABL builds dynamic baselines of application performance, and uses statistical correlations and pattern matching techniques to automatically discover performance thresholds for infrastructure components. Automated Threshold Discovery (ATD) helps by drastically reducing the efforts to configure and manually maintain performance monitoring thresholds.
ABL Analytics also provide early warnings, with deep-dive, detailed reference to infrastructure key performance indicators (KPIs) in violation. These enriched Early Warning Alerts enable faster root cause analysis (RCA) as well as automated remediation workflow.
ABL applies complex analytics algorithms to historical and real-time data sets comprised of key application performance metrics. Until now, system, database and storage administrators have executed this approach of monitoring and troubleshooting in a manual manner and “after the fact” fashion. ABL automates the complex and manually intensive analysis required to discover and troubleshoot hot spots and contention scenarios in a production environment for a highly transactional applications.
By leveraging ABL and predictive analytics, IT is now able to focus on optimizing the performance of every application, not merely the performance of the underlying device. Instead of being forced to gauge success by the number of minutes or hours to repair incidents in multi-tier environments in the traditional infrastructure-centric APM world, predictive analytics move the bar to the “left” of the chart below. IT is now able to predict and prevent issues before they occur – in some cases saving 5-10 hours to fix an incident before the outage could hit.
Manisha Arora joined Appnomic Systems as a seasoned product management vice-president empowered with a unique combination of technical analysis and interpersonal skills that allow her to communicate effectively with peers, clients and executives. Her years of business development and delivery experience demonstrate an established track record of generating significant revenue with enterprise customers. Manisha began her career a senior software engineer with Data Conversion, and then parlayed that technical experience to serve a systems analyst and senior consultant with Analyst International and Software Spectrum, respectively. Her proven expertise afforded her a transition into management consulting with Broadvision where she managed customer relationships at senior management level for Fortune 100 accounts with annual services billings of two million. Just prior to joining Appnomic, Computer Associates leveraged her technical expertise and business growth success as a solutions director, where she managed all sides of partnership, sales teams and technical architects to grow the company’s services business for deployment of Enterprise IT management solution services by 100% year over year.