Q541: Deep Java diagnostics can be low overhead
Rajagopal Marripalli , Oracle    Biography

Objectives: Monitor production Java applications with low overhead
Identify deep down application problems without any instrumentation and application knowledge
Trace transactions from Java to the Database and vice versa
Abstract: To keep your Java applications humming smoothly, you need to know how your JVM is behaving. This presentation will outline a different approach to monitoring, where you let the JVM tell you where the bottlenecks are and what application workload is responsible. By avoiding higher overhead approaches like JVMPI or bytecode instrumentation, your performance diagnosis can give you useful information without impacting your application performance. We will discuss - What do these states like Object Wait, Sleep mean. - How do I tell if I am waiting for a DB request - Can I tell what query has been sent to the database - Can I tell if I have a locking problem and find the blocker - Does the runnable state necessarily mean I am using CPU - How do I tell if the application code has been optimized

Audiernce Focus: DBA
Expertise: Advanced
Track: Middleware
Date: Tuesday, April 15, 2008
Time: 12:15 PM - 12:45 PM