Performance Tip :- Avoid Catching Exceptions

Introduction

In this post we will look at simple example how does catching exception impact performance.

Code Sample

Code for catching exception

private static void exceptionTest(){
 int i=0;
 int j=1;
 try{
 int k=j/i;
 }catch(ArithmeticException ex){
 //not good idea to catch run time exception but catch for demo only
 }
 }

Code where conditions are handled

private static void withoutExceptionTest(){
 int i=0;
 int j=1;
 if(i>0){
 int k=j/i;
 }
 }

method to call logic

private static long exceptionTestLoop(int iterations,boolean isCatchException){
 long startTime=System.nanoTime();
 for(int i=0;i<iterations;i++){
 if(isCatchException){
 exceptionTest();
 }else{
 withoutExceptionTest();
 }
 }
 long endTime=System.nanoTime();
 long time=endTime-startTime;
 return time;
 }

main method

public static void main(String[] args) {
 int itr=1000000;
 long timeForCatching=exceptionTestLoop(itr,true);
 long time=exceptionTestLoop(itr,false);
 System.out.println("Time for catching Exception "+timeForCatching+" without catching "+time);
 }

Results

exception_catch

Conclusion

Avoid catching exception as it will reduce response time. above example demonstrate results for simple single threaded application but situation become worst in multi threaded environment.

Advertisements

Compare JSON API

Introduction

In this post we will compare   two famous JSON specific API i.e.  GSON and Jackson from performance point of view. json_1

Code

Wrapper

This class is used for conversion to JSON

public class MeasurementRecord {
       private String measurementId;
      private long duration;
      private long time;
      private MeasurementType type=MeasurementType.METHOD_CALL;
       public MeasurementRecord(String measurementId, long duration, long time,
              MeasurementType type) {
             super();
             this.measurementId = measurementId;
             this.duration = duration;
             this.time = time;
             this.type = type;
       }
//getters and setters
}

code for creating list

private static List<String> getList(int iteration){
    List l=new ArrayList();
    for(int i=0;i<iteration;i++){
    l.add(new MeasurementRecord("/test.html", 10, System.currentTimeMillis(), MeasurementType.WEB_REQUEST));
    }
    return l;
    }

Jackson API

private static long jacksonTest(int iteration)throws Exception{
             ObjectMapper mapper=new ObjectMapper();
             List<String> l=getList(iteration);
             long T1=System.nanoTime();
             String json=mapper.writeValueAsString(l);
             long T2=System.nanoTime();
             return (T2-T1);
       }

Gson API

private static long gsonTest(int iteration){
             Gson gson = new GsonBuilder().create();
             List l=getList(iteration);
             long T1=System.nanoTime();
             String json=gson.toJson(l);
             long T2=System.nanoTime();
             return (T2-T1);
       }

Results results graph

Conclusion

For converting small or medium size list GSON provide better response as compared to Jackson but for large size list Jackson provide some better response than GSON. Based on this results one can conclude that for converting small or medium size list to JSON one can use GSON for better performance.

Step by step approach for web application performance improvement.

Introduction

This  post will concentrate on optimal ways to improve performance of Java based web application (although this post concentrates on Java based web application same thought process can be applied to web applications developed in other technology stack) without impacting monetary cost.

We will cover the following points from performance point of view.

  1. Important factors impacting performance of system.
  2. Patterns of performance issues
  3. Step by step approach in improving performance of system by keeping same infrastructure, i.e. without impacting cost to project.

Performance improvement activity is not a one time activity, but it is an iterative activity which involves PDCA (Plan, Do, Check, Act)

Factors impacting web application performance.

  1. Total number of users accessing (logged into) system
  2. Total number of transactions hitting to the system
  3. The Total amount of data fetched by the  client (both human and other systems)

By looking at above parameters one can easily say the performance of system degrades when any of these parameters increases.

Ways to improve performance of the system is by

  • Reducing load on system with the help of distributing load across systems (This approach mostly involves hardware cost and installation of load balancer hence we will not concentrate on this point here, but can be implemented once steps involved in this post are implemented. ).
  • Make the system work for the  least  amount of data to for request received by the client.

Patterns  for  Performance issues

Whenever a performance issue occurred one can observe any of these patterns.

  1. Issue specific to single user
    1. In order to determine RCA for these types of issue one needs to determine if client  system has incompatible software installed or some  upgradation happened recently.
  2. An issue occurred only at a specific site/location
    1. Probable root cause for these types of problems
      1. Network related issues at that specific location
    2. Incompatible software installation or some upgradation happened recently.
  3. Issue observed only for specific functionality.
    1. One needs to find the issue is specific to code, persistence layer or it is specific to external interface (webservice, JMS) system is communicating.
  4. Issue observed across all locations during a specific time frame (work hours, month end etc)
    1. These types of issues are examples of system unable to handle load.
  5. Issue observed intermittently
    1. These types of issues are more dangerous and root cause of these types of problems can be observed by taking heap/thread dumps.

Important components in  web applications

Majority of web applications will have the following components

application

Steps to improve the performance of the system.

Although one can improve performance  of system by various techniques, but ideal way is to improve performance at client layer then at application code level and at the end at persistence layer.

improve performance  from client side —>Application Code—> Persistence layer.

Steps  at   client side (browser)

These actions are related to working primarily on static content, i.e. Java Script, CSS, Images etc.

Segregate inline JS, CSS to separate JS, CSS file

Many standard browsers have important feature i.e. caching of static content (JS, CSS). Writing inline JS and CSS kills caching feature provided by the browser. Inline static content also involves pain of moving information from host to client for each client request.

This activity is a good example of making the system work on less data.

Following is screenshot for Mozilla browser indicating the amount of data transferred and the time taken in order to complete a request for first time any subsequent calls will not give call to the server but will load from the local cache of browsers.

page_loading

Replace existing JS, CSS by minified JS.

Minification is activity of removing unwanted data (whitespace, comments) from production ready static content. Few tools also provide a facility for converting big variables into small. This way one can reduce the amount of data to be transferred to client side.

A web application can have both application specific JS, CSS and  third party JS, CSS (e.g.. JQuery, Angular JS, Bootstrap…..).

Various famous  JS, CSS libraries provide both  minified and normal versions (minified version removes all unnecessary white spaces. Replace long identifiers by short).

Following is screenshot indicating the amount of data downloaded and time taken for loading minifed version of angular js.

page_loading_min

The total amount of improvement in time  (509-78) *100/509 nearly 84%

Total amount of improvement in data (937.81-122.96)*100 /937.81 nearly 86%

Compressing application specific static content

In order to compress application specific static content (JS, CSS) various modification tools available

Few common examples of compressors

  1. YUI Compressor: – Tool from Yahoo
  2. JsMin: – from Douglas Crockford, one of the most respected personality in the field Java script.
  3. Google Closure: -Tool from Google

Remove Http 404 status code

Http 404 status code sent by the http host  when requested resource not available to host. These types of requests create unnecessary network round trips.

Following links help in configuring access logs for jboss and Weblogic server.

Configure access logs in Weblogic

Configure access logs in JBOSS

Imp Note: - enabling access log in, the system will increase load on system, hence must be used with caution.

Replace client side pagination/huge table data by server side pagination

Server side pagination help loading page quickly. This technique fetches only required records to be displayed on the view.

Another way to improve page loading time is by forcing users to narrow down searches.

Steps  on   the server  (application/web server) side

Integrate application with the application performance monitoring (APM) tool.

Purpose of integrating these tools is to measure performance of the system in production with minimal load on system. It also helps in determining parameters during load. Although the majority of these tools utilizes java’s bytecode instrumentation, but JAMon provides Aspect oriented programming (AOP).

Important parameters from performance point of view.

  • Maximum number of users logged into the system.
  • Maximum transactions hitting to the system (determine transaction rate)
  • Most frequently hit URLS on the system.
  • Slow Running transactions in the system.
  • Slow Running SQL queries.
  • Most frequently occurred Exceptions.

Following are few important APM tools.

table

Some other APM tools are New Relic, Compuware Dyna trace. These tools  along with AppDynamics present in Gartner’s magic quadrant.

Reduce IO

One can use following techniques for reducing IO.

  • Reduce number of logs to file
  • Remove System.out.println
  • Reduce size of log file :- it is always better to write to small files instead of writing and analyzing huge files
  • Avoid catching too many exceptions in the system. Handle it properly. Following code indicates how system will behave for exception and without exception. Various APM tools provide most frequently thrown exceptions and code which is throwing them.
public class ExceptionTest {

       private static void exceptionTest(){

             int i=0;

             int j=1;

             try{

                    int k=j/i;

             }catch(ArithmeticException ex){

             }

       }

       private static void withoutExceptionTest(){

             int i=0;

             int j=1;

             if(i>0){

                    int k=j/i;

             }

       }

       private static void exceptionTestLoop(int iterations,boolean isHandleException){

             long startTime=System.nanoTime();

             for(int i=0;i<iterations;i++){

                    if(isHandleException){

                           exceptionTest();

                    }else{

                           withoutExceptionTest();

                    }

             }

             long endTime=System.nanoTime();

             long time=endTime-startTime;

             System.out.println(“time for execution  exception “+time+” iterations “+iterations+” isHandleException “+isHandleException);

       }

       public static void main(String[] args) {

             exceptionTestLoop(100000,true);

             exceptionTestLoop(100000,false);

       }

}

results

Steps  on the DB side.

Although this post concentrate mainly on application code and browser specific side but following are easier techniques on persistence layer.

  1. Add indexes to columns: – this technique improves performance of Select Queries but reduces DML hence must be used with caution.
  2. Use run time parameters to SQL queries instead of using hard coded static parameters
  3. Cache master data instead of fetching it from db
  4. Archive database on predefined interval
%d bloggers like this: