Hi,
Well, I think we've done a sort of similar project,
and I would just like to discuss the strategy we've
adopted. I am not sure whether it is really efficient
or not, but it works..
I've used JSP pages as the view, Java Beans to
encapsulate the business logic, and the database was
MySQL.
Whenever a User logs in, a fresh Bean is created for
him. This would remain in the server as long as that
session is active. This bean maintains the state of
the User throughout. The advantage of this approach
is that it gives us the flexibility of future
scalability, without in any way disturbing the view
and the underlying database.
So, each User Bean is at Session scope. There is one
bean at Application level, which stores the references
to all the active User BEans. This enables us to
cross check, whenever a user tries to log in , whether
he is already active or not. If yes, access is not
allowed, thereby preventing the logging of user twice.
When any session ends, the reference is nullified and
the Bean is destroyed.
Another application level Object, in which the
Connection Pooling is implemented, is available for
all the User Beans. All the User Beans get a free
Connection returned by a method in this Object,
execute the query and fetch the resultset and return
back the connection.
I've implemented my own Connection Pooling class and
also a third party ConnectionBroker class. Both are
working perfectly fine.
Pls tell me the disadvantages, if u suspect, in the
above design.
Y.Laxminarayana
--- Rishi Shehrawat <[EMAIL PROTECTED]>
wrote:
<HR>
<html><DIV id=cdiv>
<P>Hi, <BR><BR>I know probably this has been discussed
on the forum before, i tried finding out some answers
to my question but was unable to do so. <BR><BR>I will
try explaining the problem <BR>We are in the process
developing a portal for a client. <BR>This is
basically a search based site with absolutely no
transactions although they might be added later when
the site is e-commerce enabled. <BR>There are no dbase
updates <BR>There are different categories, &
sub-categories & surfers can get search results
based on the search criteria given by them. <BR>There
is a generic keyword search & a
category/sub-category specific search where a a surfer
can get the results based on the narrowing options for
the specific category/sub-category. <BR>The client
will most probably be using IBM Websphere application
server & NES web server. <BR>Time taken to process
a client request is very important <BR><BR><BR>The
different options availiable to us are (to the best of
m!
y knowledge): <BR>Entity Beans <BR>Stateful &
stateless Session Beans <BR>Servlets/JSP's <BR><BR>I
think entity beans are out as there are no
transactions & security issues. They would take up
a lot of resources. I was also thinking that stateful
session beans would not be a very good idea. I think
it would be best to perform the search & keep the
results as a session object instead of having a
stateful session to service the request. Now we are
left with stateless session beans & servlets.
<BR><BR>The stateless beans will have the SQL
statements. From what i have read a few stateless
session bean beans can request a lot of clients. As
mentioned in the Gemstone Developer's guide a pool of
instances can be maintained at the client servlet so
that i can save the lookup & create time .Also the
database connection pooling will again help in cutting
down on the processing time. Plus the other advantages
of the application server. Caching of results is
another !
option but i am not very sure if it will work in our
case as there are different queries for different
categories, a single cache cannot sevice the requests
& there are a large number of the
categories\sub-categories to write a cache specefic to
each category. <BR>The immediate downside i could
think up of was the network traffic between the web
server & the application server. <BR><BR>In case
of a servlet i would directly fire a SQL get the
results from the dbase & service the clients
request. I bypass the app server. <BR>The problem i
see here is the load handling capabilities of the web
server. The web server might not provide facilities
like connection pooling.The presentation layer will be
tightly coupled with the business logic. I could
probably overcome this by using java objects.. But
this could again slow down the system... <BR><BR>So as
you can see i am quite confused on which would be a
better approach. I feel that SSB's might be a better
approach, but!
again i am not very sure !!! Please let me know your
experiences. Which approach do you think is best
suited for the requiements mentioned above.
<BR><BR>Thanks In Advance <BR>Rishi
<BR></P></DIV><p><hr>Get Your Private, Free E-mail
from MSN Hotmail at <a
href="http://www.hotmail.com/">http://www.hotmail.com</a><br></html>
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED]
and include in the body
of the message "signoff EJB-INTEREST". For general
help, send email to
[EMAIL PROTECTED] and include in the body of the
message "help".
__________________________________________________
Do You Yahoo!?
Kick off your party with Yahoo! Invites.
http://invites.yahoo.com/
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
