Hi Ville,
Do you, by any chance, have different names defined for your controller
servlet in the web.xml file? I also have my own ControllerServlet
framework and I just have a single instance, unless I define more than
one name for the same servlet. If you define different web applications
you would also have different instances, even if you use the same name,
so have a look at your web.xml files to see if the problem is in there.
Just my 2ec,
D.
Ville Rinne wrote:
>
> Kevin, let me know how your tests came out. I did some more testing myself
> and I found it strange that when I created a specific simple test-servlet
> for this case I didn't manage to get more than one instance of it created.
> I'll do some more testing tomorrow but there is definitely something
> strange going on. Obviously the amount my testservlet takes to complete a
> request is smaller than the amount of time my request-controller takes but
> at this point I honestly don't have a clue :P
>
> Anyhow, since servlets aren't synchronized in any way by default, how
> would it increase performance even if there was more than one instance of
> them in existance?
>
> cheers,
> Ville Rinne
>
> Of course it isn't a huge problem
>
> On Tue, 7 Aug 2001, Duffey, Kevin wrote:
>
> > Keep in mind that every app server is allowed to implement the spec in their
> > own manner. It does sound peculiar, but it is very possible that Orion gets
> > its performance by pooling a few servlet instances even without the single
> > threaded model. I don't know for sure, but that seems logical. By having
> > even two or three instances, it could increase the ability to handle ore
> > requests..although myself I thought that a single instance would be plenty
> > fast to have literally thousands of requests per second hit it. Each time a
> > request comes in, the container should create a new thread for that request
> > then send it on its way to the servlet instance. A PIII600 system can run
> > over 1billion instructions per second. Figure that with the JVM interpreter,
> > your code and any resources, you should still be able to iterate thousands
> > of requests per second, unless object creation in a servlet (and depending
> > on what it is doing) slows things down. The one factor I am not sure of is
> > how many "threads" a JVM can handle at one time. I have heard of single
> > servers handling 1000's of hits per second, and if each request goes to a
> > JSP or servlet, that should equate to one new thread per request..even if
> > the session id is the same (for example, the user somehow opened a new
> > window from the browser window..and could possibly submit two (or more)
> > requests at the same time).
> >
> > At any rate, it sounds to me like perhaps either there is a bug, or that
> > Orion is doing this pooling automatically to increase performance.
> > Interesting though..I will now check out to see if my MVC framework does the
> > same thing.
> >
> >
> >
> > -----Original Message-----
> > From: Ville Rinne [mailto:[EMAIL PROTECTED]]
> > Sent: Tuesday, August 07, 2001 7:59 AM
> > To: Orion-Interest
> > Subject: multiple instances of a servlet?
> >
> >
> > We've implemented the request-controller pattern using a servlet as the
> > entry-point of all our jsp-pages. For some reason Orion creates multiple
> > instances of this servlet even though it isn't implementing
> > SingleThreadModel and there isn't anything peculiar about it. I have the
> > servlet printing to System.out every time init() is run and every time it
> > receives a request, it also prints the memory address of the instance that's
> > handling it. For some reason with this servlet there are multiple instances
> > (up to 3 so far) that handle these requests. There doesn't seem to be any
> > larger logic in which instance gets to handle the request. It's not a
> > problem per se but I just find it rather strange since I assumed that only
> > one instance of a servlet would be created. Anyone else ran into this ?
> >
> > cheers,
> > Ville Rinne
