Introduction

Do not store large object graphs in HttpSession. This can severely affect the performance of a WebSphere application. Look for alternatives like an application specific JDBC framework that can augment or replace HttpSession for storing servlet state data. Application design for performance and scalability requires consideration of servlet state data. If your application requires a large amount of servlet data, consider application specific alternatives that either augment or replace the use of HttpSession.

This best practice applies to the following product, version, and platform:

• WebSphere Application Server Base, all platforms, versions 3.0.2.x, 3.5.x, 4.0

Back to top

Recommendation

Large applications require using persistent HttpSessions. However, there is a cost. An HttpSession must be read by the servlet whenever it is used and rewritten whenever it is updated. This involves serializing the data and reading it from and writing it to a database. In most applications, each servlet requires only a fraction of the total session data. However, by storing the data in the HttpSession as one large object graph, an application forces WebSphere® Application Server to process the entire HttpSession object each time.

WebSphere Application Server has HttpSession configuration options that can optimize the performance impact of using persistent HttpSessions. The HttpSession configuration options are discussed in detail in the WebSphere Application Server documentation. Also consider alternatives to storing the entire servlet state data object graph in the HttpSession. Using your own JDBC connection alternative is described and discussed below.

Do not forget to explicitly invalidate the HttpSession when you are finished with it.

When configuring persistent sessions in WebSphere Application Server, use a dedicated data source. To avoid contention for JDBC connections, do not reuse an application DataSource or the WebSphere Application Server repository for persistent session data.

Figure 1 compares the relative performance of a sample application with a single object of different sizes. As the size of the objects stored in the HttpSession increases, throughput decreases, in large part due to the serialization cost.

Back to top

Alternative

In most applications, each servlet needs only a fraction of the entire application's state data. As an alternative to storing the entire object graph in the HttpSession, use JDBC for partitioning and maintaining the state data needed by each servlet in the application.

A sample JDBC solution maintains the state data needed by each servlet as a separate row in an application-maintained JDBC datasource. The primary keys for each row (the data for each servlet) are stored as separate attributes in the HttpSession. The use of the HttpSession is reduced to the few strings needed to locate the data.

JDBC alternative - session data

// Get the session data value
try {
           conn= getPooledConnection(); // Uses DataSources to Get JDBC Connection 
           	See Best Practice
           conn.setAutoCommit(false);
           sessionKey = (String) session.getValue(TOP_KEY);

           // Session Key Does Not Exist - This is a new Session Build the Data
           if (sessionKey == null) {
                      sessionKey = UniqueValue.getUniqueValue();
                      session.putValue(TOP_KEY, sessionKey);
                      sessionDtaa = new SessionLargeObjectCollection(NUM_REPS);

                      psi = conn.prepareStatement(INSERTSTMT);
                      psi.setString(1,sesisonKey);
                      psi.setBytes(2,convertToBytes(sessionData)); // Serialize and Write Out
                      psi.executeUpdate();

                      sessionKey = UniqueValue.getUniqueValue();
                      session.putValue(Integer.toString(i+1),sessionKey);
           }
           else { // The Session Does Exist...Retrieve the Data to Generate Load
                      sessionKey = (String)session.getValue(TOP_KEY);
                      pss.conn.prepareStatement(SELECTSTMT);
                      pss.setString(1,sessionKey);
                      rs.pss.executQuery();

                      if (rs.next()) { // De-serialze and read in
                                   sessionData = convertBytesToSessionObjectCollections
                                       (rs.getBytes("SAVESERIALIZEDDATA"));
                      }

                      // Update the Data to Generate Load
                      psu = conn.prepareStatement(UPDATESTMT);
                      psu.setString(2,sessionKey);
                      psu.setBytes(1,convertToBytes(sessionData)); // Serialize and Write Out
                      psu.executeUpdate();
           }
           conn.commit();
}
catch (Exception e) { // Handle Errors }
finally {
           // Close Connections and Statements - See Best Practice
}

Figure 2 shows the performance improvement of partitioning the servlet state data and using JDBC over maintaining servlet state data as a single object in the HttpSession.

JDBC alternative - cleanup Issues

The homegrown servlet state data framework illustrated in the example creates records in a user database. At some point, these records need to be deleted through a cleanup process. Two good alternatives for this purpose are:

• Periodic offline: Non-WebSphere process. On UNIX systems, this would be a CRON job. On Windows® NT or Windows 2000, it would be a scheduled function. In either case, it would delete records after a certain period of time depending on the application. In this example, the database is keyed on servlet data time of creation.
• HttpSessionBindingListener: This homegrown alternative stores the servlet state data in an application - specific database. It stores the primary keys for the servlet state data records in the HttpSession. When the HttpSession is about to be destroyed, the javax.servlet.http.HttpSessionBindingEvent for valueUnbound could direct a delete of the servlet state data. If the lifetime of servlet state data is longer than that of the HttpSession, use the first alternative instead.

Resources

• WebSphere Application Server Development Best Practices for Performance and Scalability
  
  

About the author

Rate this article

Comments

Back to top

Table of contents

• Introduction
• Recommendation
• Alternative
• Resources
• About the author
• Comments

Next steps from IBM

Rational Application Developer provides a comprehensive Integrated Development Environment (IDE) used to develop Web 2.0 applications requiring J2EE, Servlets and JavaServer Pages (JSPs).

---

• Try: Try the free Rational Application Developer (RAD), which helps developers quickly design, develop, test, analyze, and deploy high-quality Java, Javaserver pages, J2EE and Web 2.0 applications.
• Article: See how Rational Application Developer and Service Data Objects have been extended to JavaServer Pages to create data views and database queries using a simple, wizard-based, drag-and-drop technology.
• Tutorial: Take a look at developing a J2EE Web application and JavaServer Pages (JSPs) for both an HTML input form and the HTML output from the server, using Rational Application Developer.
• Demo: This demo highlights the enhanced Web 2.0, rich Internet application development features in Rational Application Developer including AJAX and Java Persistence API application development.
• Buy: Rational Application Developer for WebSphere Software

Dig deeper into WebSphere on developerWorks

IBM SmartCloud trial. No charge.

Unleash the power of hybrid cloud computing today!

---

Public and private solutions in one trial.

Special offers