Minimize the use of synchronization in servlets. Because servlets are multi-threaded, synchronization of the major code path can seriously and adversely affect performance.
Servlets are multithreaded. Servlet-based applications have to recognize and handle this appropriately. If large sections of code are synchronized, an application effectively becomes single threaded, and throughput decreases dramatically.
No synchronization in servlets presents the best option, however, if the application design cannot avoid synchronization, then use a "Lock Object" and lock the smallest possible code path. Do not synchronize the servlet service method or the doGet and doPost methods. These methods are the major code paths. Synchronizing these methods or any of the servlet methods will lock the entire servlet instance. The following code shows an example using a "Lock Object" to protect the servlet instance variable numberOfRows.
Minimum synchronized code path
ublic class BpAllBadThingsServletsV1b extends HttpServlet { private int numberOfRows = 0; private javax.sql.DataSource ds = null; private Object lockObject = new Object(); public void doGet(HttpServletRequest request, HttpServletResponse response( throws ServletException, IOException { Connection conn = null; ResultSet rs = null; PreparedStatement pStmt = null; int startingRows = 0; synchronize(lockObject) { startingRows = numberOfRows; } try { String employeeInformation = null; conn = ds.getConnection("db2admin", "db2admin"); pStmt = conn.prepareStatemtn ("select * from db2admin.employee"); rs = pStmt.executeQuery(); } catch (Exception es) { // Error handling code here } } }
The following code shows how the major code path is synchronized to protect a servlet instance variable called numberOfRows.
Using javax.servlet.SingleThreadModel is yet another way to protect servlet instance variables, but this should be avoided as well.
Figure 1 shows the performance impact of Synchronization
Locking the major code path: excessive synchronization
public class BpAllBadThingsServletsV1a extends HttpServlet { private int numberOfRows = 0; private javax.sql.DataSource ds = null; public void doGet(HttpServletRequest request, HttpServletResponse response( throws ServletException, IOException { Connection conn = null; ResultSet rs = null; PreparedStatement pStmt = null; int startingRows; try { synchronized(this) // Locks out Most of the Servlet Processing { startingRows = numberOfRows; String employeeInformation = null; conn = ds.getConnection("db2admin", "db2admin"); pStmt = conn.prepareStatemtn ("select * from db2admin.employee"); rs = pStmt.executeQuery(); } } catch (Exception es) { // Error handling code here } } }