Java Performance : Measuring and Tuning Nataraja Neelakanta - PowerPoint PPT Presentation

Java Performance : Measuring and Tuning Nataraja Neelakanta

Overview Most of the applications today are designed • and developed for simultaneous use by a large number of users. At times there are underlying problems in • the code, which may cause error conditions the code, which may cause error conditions such as Memory leaks, Race conditions, Thread locks and so on which if neglected, stretches the use of resources beyond limits, causing the application to slow down or even crash.

Profiling Java Profiling is the process of monitoring various JVM level parameters such as Method Execution, Thread Execution, Object Creation and Garbage Collection. CPU time being utilized per method CPU time being utilized per method • Memory being utilized • Method call information • Objects being created • Objects being garbage collected •

Profile snapshot example

JRE / JVM hmm … what is what ?

More on JRE and JVM JIT converts the .class file to machine • executable code that can be directly executed by OS. This machine executable code is placed in • class segment of RAM by class loader. class segment of RAM by class loader. Java <Program Name> starts the jvm, which • in turn invoke MM module of JRE to allocate three segments in RAM, based on -XX parameters. JRE can be looked upon as an instance of • java.lang.Runtime class

Classloaders : hierarchy and functions

Class loading phases

Class loading explained Physical loading- class file will be searched 1. in the classpaths, if found, the byte code is loaded. A basic memory structure is established hence. Linking – Linking – 2. 2. i. Byte code verification 1. ii. Class preparation – prepares data structures 2. fields, methods etc. iii. Resolving references. 3. 3. Initialization – static blocks, variables.

Understanding the heap The Java heap is divided into three sections • - Young generation, Old generation, Permanent Generation. HotSpot VM • The HotSpot VM is the garbage collector that comes with the JavaSoft virtual machine. The heap layout of a typical HotSpot VM is • depicted in the next slide

Heap layout

Young Generation The Eden Space of the Young Generation • holds all the newly created objects. When this generation fills,the Garbage Colle • ctor clears out of memory all objects that are unreferenced. unreferenced. Objects that survive this gcare moved to the • “From” Survivor Space. It has two equally ‐ sized subspaces “To” and • “From” which are used by its algorithm for fast switching and cleanup.

Old generation Once an object survives a given number GC • ,it is promoted (or tenured) from the “To” Sp ace to the Old Generation. Objects in this space are never garbage coll • ected except in the two cases- ected except in the two cases- Full Garbage Collection • Concurrent Mark ‐ and ‐ Sweep Garbage Collection • If the Old Generation is full and there is no way for the heap to expand, an Out ‐ of ‐ Memory error (OOME) is thrown and the JVM will crash.

Permanent Generation The Permanent Generation is where class fil • es are loaded/kept. These are the result of c ompiled classes and jsp pages. If this space is full, it triggers a Full Garbage • Collection. Collection. If the Full Garbage Collection cannot clean • out old unreferenced classes and there is no room left to expand the Permanent Space, an OOM is thrown and the JVM will crash.

Code snippet causing PermGen

Memory image for the class

Reachability and GC

Performance improvement – General aspects Using appropriate data structures. 1. Making use of multi-threaded approach 2. Using WeakHashMaps as appropriate. 3. Careful usage of Singletons. Careful usage of Singletons. 4. 4. Removing unnecessary methods, variables, 5. loops. Avoiding multiple calls to remote 6. components. Implementing efficient hashing algorithms. 7.

Reference utility-java.lang.ref Garbage Collector won’t remove a strong reference . • A soft reference will only get removed if memory is low. So • it is useful for implementing caches while avoiding memory leaks. A weak reference will get removed on the next garbage • collection cycle. Can be used for implementing canonical collection cycle. Can be used for implementing canonical (conformance to standards)maps. The java.util.WeakHashMap implements a HashMap with • keys held by weak references. A phantom reference will be finalized but the memory will • not be reclaimed. Can be useful when you want to be notified that an object is about to be collected.

Performance improvement – across layers Plain java layers – Effective use of java.util.concurrent 1. package utilities like ConcurrentHashMap. Future interface can be used to represent 2. results of an asynchronous computation. results of an asynchronous computation. Using UncaughtExceptionHandler to get 3. notified when a thread is about to get terminated due to uncaught exception. If UEH is used along with an executor 4. service, care should be taken to use it via Future. Making thread a daemon. 5.

Performance improvement – across layers…Plain java layers Plain java layers – 5. Using of BlockingQueue, especially where there is a need of Producer /Consumer scenario. 6. Effective usage of ThreadPoolExecutor, that provide improved performance when executing large numbers of asynchronous tasks, due to reduced per-task invocation overhead. 7. They provide a means of bounding and managing the resources, including threads, consumed when executing a collection of tasks. 8. Effective String handling.

Performance improvement – across layers…J2EE stack 1. Using AJAX, that cuts the response and presents only relevant part of response to the client. 2. Effective design and implementation, using responsive xml parsing mechanisms, like responsive xml parsing mechanisms, like StAX. 3. Using JSON for presentation tier, for data response, which gets interpreted faster than xml. 4. Using Proxy / Prototype patterns wherever necessary, whenever object creation / initialization is costly.

Performance improvement – across layers…J2EE stack Database interaction Using Prepared (pre-compiled) Statements 1. Using frameworks such as Spring, 2. Hibernate. Ensuring closing of Connection, ResultSet. 3. Being rational in using Prepared vs normal 4. Statements. Reducing Data Intensive operations in Java 5. layer. Using frameworks such as quartz for job 6. scheduling.

Performance improvement – across layers…J2EE stack Databases / Infrastructure Using Indices – Clustered / Non-Clustered. 1. Denormalization of schema 2. Query hints. Query hints. 3. 3. Optimal usage of joins, and avoiding 4. inner/subqueries. Fail-over, load balancing mechanisms 5. offered by application servers. Using cloud, SaaS,PaaS,IaaS. 6.

Performance improvement – across layers…J2EE stack Databases / Infrastructure Sharding Horizontal partitioning is a database design principle • whereby rows of a database table are held separately, rather than being split into columns. Each partition forms part of a shard , which may in turn be located on a separate database server or physical location. Since the tables are divided and distributed into multiple servers, • the total number of rows in each table in each database is reduced. This reduces index size, which generally improves search • performance. A database shard can be placed on separate hardware, and • multiple shards can be placed on multiple machines. Sharding differs from horizontal partitioning, where, Sharding has • the capability to partition across multiple instances of the schema. Hibernate supports shards. •

Conclusion Java Performance is an oceanic topic There should be an optimal / rational i. design. More time should be spent on application ii. design, and preparation for implementation. Preparation for implementation means, iii. finalizing the data structures, data types Performance measurement should be part iv. of the project plan and a continued effort.

Thank you !! Thanks to all for your patience !