Java Performance Monitoring Libraries

There is a proposal to build performance probes in WSO2 Platform. For that I started looking in to some performance monitoring libraries.

Following libraries were mentioned in the WSO2 architecture thread.
While looking in to these libraries, I found out about following also.

Here is a quick comparison of each project. These comparison criteria are based on the requirements in above proposal.



Metrics ParfaitJAMonJava SimonPerf4J
LicenseApache License 2.0Apache License 2.0JAMon LicenseNew BSD LicenseApache License 2.0
SourceGitHubGoogle CodeSourceforgeGitHubGitHub
Latest Version3.1.00.2.82.794.0.00.9.16
Last PublishedSep 4, 2014Jun 01, 2011Aug 20, 2014Oct 29, 2014Oct 16, 2011
Java Version-Java 6-Java 7-
CountsYesYesYesYesNo
TimingsYesYesYesYesYes
JMX SupportYesYesNoYesYes
Enable/DisableNoNo*YesYesYes*

* Not confirmed

Let's look at each library in brief.

Metrics


Metrics provides various measuring instruments. 
  • Meters - Measuring rate of events over time
  • Gauges - Instantaneous measurement of a value
  • Counters - Measurement for counting
  • Histograms - Statistical distribution of values
  • Timers - Measures the duration of a code block and the rate of invocation.
  • Health Checks - Centralizing the health checks of services

Metrics has modules for common libraries like Jetty, Logback, Log4j, Apache HttpClient,
Ehcache, JDBI, Jersey.

Metrics provides a way to have multiple reporting options. Mainly JMX, Servlets (http), Console, CSV and SLF4J.  It also supports Ganglia and Graphite reporting

Metrics' Getting Started page shows you how to use the Metrics APIs.

Parfait


Parfait provides mechanisms for collecting counter and timing metrics. Data can be exposed via various mechanisms including JMX and the the open-source cross-platform Performance Co-Pilot.

Parfait also has number of modules, which enable to collect metrics from common data sources.

JAMon


JAMon provides various ways to monitor applications without code changes. It has in built support for HTTP Monitoring, Spring, JDBC, Log4j, EJB etc. JAMon can be used in Servlets, JSPs, EJBs and Java Beans in various Java EE Application Servers.

JAMon doesn't seem to support JMX.

Java Simon


Java Simon has monitors called Simons, which can be used in the code to count something or to measure the time taken.

It's interesting to know that the Java Simon was started by the people, who used JAMon earlier. They were not satisfied with JAMon in terms of simplicity and monitor structure. Some people also consider Java Simon as the replacement of JAMon.

Java Simon also measures time in nanos. Simons are organized in a hierarchy.

Simons can be disabled easily. 

Java Simon has a web console in addition to exposing data via JMX. There are many examples for Java Simon usage and Getting Started wiki is a good place to see how we can use the APIs.

There is comparison of Java Simon with JAMon, which shows performance overhead of each library.

Perf4J


The Perf4J mainly makes use of the logging frameworks and it has support for popular logging frameworks.

The Perf4J support is limited only to timing metrics and I didn't find the support for counters.

Summary


In this blog post, I just wanted to give an idea about Java Performance Monitoring libraries and each library has pros & cons. So, depending on the project requirements, it's better to evaluate all libraries and select a suitable one for your project.

There is lot of information in each project's web pages. Going through those pages will help to understand more about features provided by each library.

Location: Kollupitiya, Colombo, Sri Lanka

Comments

Post a Comment

Popular posts from this blog

Specifying a custom Event Settings file for Java Flight Recorder

Image
When you are using Java Flight Recorder (JFR), the JFR will use an event settings file to check which event types to record. By default in JFR, there are two settings, "default" and "profile". The default setting is recommended for Continuous Recordings as it has very low overhead (typically less than 1% overhead). The profile setting has more events and useful when profiling the application. As mentioned in my previous blog post regarding  Java Flight Recorder Continuous Recordings , we use following arguments to do a Continuous Recording. -XX:+UnlockCommercialFeatures -XX:+FlightRecorder -XX:FlightRecorderOptions=defaultrecording=true,disk=true,repository=./tmp,dumponexit=true,dumponexitpath=./ Note: According to the Oracle documentation on " java " command, we should be able to specify "settings" parameter to the -XX:FlightRecorderOptions. However, the settings parameter has no effect when used with the -XX:FlightRecorderOptions and
Read more

Flame Graphs with Java Flight Recordings

Flame Graphs Brendon D. Gregg , who is a computer performance analyst , has created  Flame Graphs to visualize stack traces in an interactive way. You must watch his talk at USENIX/LISA13 , titled Blazing Performance with Flame Graphs , which explains Flame Graphs in detail. There can be different types of flame graphs and I'm focusing on  CPU Flame Graphs  with Java in this blog post. Please look at the Flame Graphs Description  to understand the Flame Graph visualization. CPU Flame Graphs and Java Stack Traces As  Brendon  mentioned in his talk, understanding why CPUs are busy is very important when analyzing performance.  CPU Flame Graphs  is a good way to identify hot methods from sampled stack traces. In order to generate CPU Flame Graphs for Java Stack Traces , we need a way to get sample stack traces. Brendon has given examples to use jstack  and Google's lightweight-java-profiler . Please refer to his perl program on generating  CPU Flame G
Read more

Benchmarking Java Locks with Counters

Image
These days I am analyzing some Java Flight Recordings from taken from WSO2 API Manager performance tests and I found out that main processing threads were in "BLOCKED" state in some situations. The threads were mainly blocked due to "synchronized" methods in Java. Synchronizing the methods in a critical section of request processing causes bottlenecks and it has an impact on the throughput and overall latency. Then I was thinking whether we could avoid synchronizing the whole method. The main problem with synchronized is that only one thread can run that critical section. When it comes to consumer/producer scenarios, we may need to give read access to data in some threads and write access to a thread to edit data exclusively. Java provides ReadWriteLock for these kinds of scenarios. Java 8 provides another kind of lock named StampedLock . The StampedLock provides an alternative way to the standard  ReadWriteLock  and it also supports optimistic
Read more