Load Tests for (Desgin issue, Servlets with EJBs in real life?)

Question

I have a very light java program that makes an HTTP connection to a servlet
on the WLS,
it uses the GET method for that, and reads the response from the server, it
logs the time that passes from just before opening the HTTP connection and
right after finishing reading the response. Below is the code snippet for
that:
url = new URL(sURL);
start = getCurrentTime();
URLConnection conn = url.openConnection();
conn.connect();
readResponse(conn.getInputStream());
end = getCurrentTime();
There is a second piece that controls the HTTP connections, it is a separate
java program that creates
the HTTP connection process as many times as desired in a loop and waits for
a specific period of time between every fork of the process (values I used
are 0ms, 3 ms, 30ms etc), you can think of this as arrival rate for the
requests to the server. Below is the code snippet that does this :
for(int i=0; i<nNumOfHits; i++) {
Runtime.getRuntime().exec(HTTP_conn_command);
Thread.sleep(nInterval);
I used WLS 5.1, on a pentium III 866 Mhz., Windows 2000 Pro, with 512 Mb
Ram,
tests are done locally, (WLS and the client programs run on the same
machine),
after each test I refresh the server to be accurate.
With the Servlet + Java approach what the servlet does is to instantiate a
Java Object and
call an Init method on the object passing the HttpServletRequest to it, and
call the
corresponding setters, that's it. There are 7 paramaters which are all
strings, so 7 setters are called.
The Servlet + EJB approach does the same thing, but encapsulates the above
implementation in a single business method of a stateless session bean. Home
interface is created in Init() method of the servlet,
so it is used over and over again.
Any comments on this, greatly appreciated,
thanks,
ali.

Answer 1

I have a very light java program that makes an HTTP connection to a servlet
on the WLS,
it uses the GET method for that, and reads the response from the server, it
logs the time that passes from just before opening the HTTP connection and
right after finishing reading the response. Below is the code snippet for
that:
url = new URL(sURL);
start = getCurrentTime();
URLConnection conn = url.openConnection();
conn.connect();
readResponse(conn.getInputStream());
end = getCurrentTime();
There is a second piece that controls the HTTP connections, it is a separate
java program that creates
the HTTP connection process as many times as desired in a loop and waits for
a specific period of time between every fork of the process (values I used
are 0ms, 3 ms, 30ms etc), you can think of this as arrival rate for the
requests to the server. Below is the code snippet that does this :
for(int i=0; i<nNumOfHits; i++) {
Runtime.getRuntime().exec(HTTP_conn_command);
Thread.sleep(nInterval);
I used WLS 5.1, on a pentium III 866 Mhz., Windows 2000 Pro, with 512 Mb
Ram,
tests are done locally, (WLS and the client programs run on the same
machine),
after each test I refresh the server to be accurate.
With the Servlet + Java approach what the servlet does is to instantiate a
Java Object and
call an Init method on the object passing the HttpServletRequest to it, and
call the
corresponding setters, that's it. There are 7 paramaters which are all
strings, so 7 setters are called.
The Servlet + EJB approach does the same thing, but encapsulates the above
implementation in a single business method of a stateless session bean. Home
interface is created in Init() method of the servlet,
so it is used over and over again.
Any comments on this, greatly appreciated,
thanks,
ali.