New AppDynamics Software Hunts Memory Leaks, Finds Root Cause, and it's All in Production

Java memory issues are common and often difficult to diagnose.   Profilers and other tools are great, but they have their blind-spots.  For production environments, profilers can constitute a lot of overhead.  They rely on heap dumps instead of runtime data, and the heap dump approach is not suitable for large heap sizes that are commonly found today.  Some profilers have non-heap dump approaches, but they only capture shallow object sizes.  This was AppDynamics' view of the production memory monitoring and diagnostics space when they decided to get into the game.  With today's release of AppDynamics 3.0, they're showing companies the value of a new approach - memory leak detection and root cause diagnostics in the production environment.

“Memory leaks create havoc for countless organizations with mission-critical java applications,” said AppDynamics CEO Jyoti Bansal. “Best case scenario, a memory leak causes your system to slow down, dragging application performance well below established SLAs. Worst case scenario, your servers crash completely and you don’t know why."

Here is AppDynamics' ideal memory troubleshooting flow:



You can find legacy tools that identify memory leaks pretty effectively, but AppDynamics says that their release today is the first that also provides analytics that can determine the root cause of the leaks within the production environment.  This is important because more and more companies want to evaluate their caching strategy.  

AppDynamics 3.0 enables real-time Java heap monitoring, garbage collection memory pool monitoring, an shows the correlation between the heap and the major and minor GC collections:



Root cause diagnostics in AppDynamics 3.0 will look at code paths and transactions and determine which ones are accessing the collection.  Bansal lists some of the other distinguishing features of AppDynamics:

• Low-cost algorithm for object deep size calculation
• Automatic Java collection instrumentation
• Dynamic access/allocation code path analysis
• Live object instance tracking
• Automatic memory leak detection with best-fit linear regression analysis

AppDynamics 3.0 also builds on its feature set for highly distributed cloud environments.  Bansal asserts, “Cloud applications require a performance management solution that can dynamically discover, map, instrument and monitor the environment - even when 100 nodes appear or disappear
on the fly.”

3.0 includes a new dynamic cluster aggregation feature that uses self-learning baselines to intelligently track instance lifecycles.  The new version also facilitates up to 1000 cloud node agents reporting to a single AppDynamics controller.  Finally, 3.0 includes dozens of customer-requested features: