Re: queue system

From: Federico (es_at_yahoo.fedeabascal)
Date: 09/03/04

  • Next message: Federico: "Re: queue system" 
    Date: Fri, 03 Sep 2004 13:42:26 +0200

    Uff... thanks a lot guys. How much information.

    What I want is very similar to what Christopher suggested, but the
    problem is that the programs are waiting until the program finishes, but
    then how could I now which program is next?

    I explain it better: the user sends information to my server and I run a
    program that can take some hours. I notify the user that when his work
    is finished he will receive an email. Only one copy of the program
    should run at a given time. If more users send request for running it,
    and it is already running, they should wait (well, not the user but the
    requests).

    I programmed something similar some time ago: the program requests were
    appended (writen) to a file, and when a program finishes, it removes its
    request from the top of the file (locking while those file operations
    were carried out, and putting waiting programs to sleep for 5 minutes).
    But I was thinking it will be better to use a better (standard, already
    built) queue system. If I had so, I could say: there can be two programs
    running at once instead only one, for example.

    Well... thanks again,
    Federico

    P Gentry wrote:
    > Federico <es@yahoo.fedeabascal> wrote in message news:<ch75cv$bjc$1@nsnmrro2-gest.nuria.telefonica-data.net>...
    >
    >>Hello, people
    >>
    >>I have to build a web page for running a program. But I want no more
    >>that one copy of the program running at a given time: I want a queue
    >>system. Can anyone suggest which one could I use? or where to find this
    >>information?
    >>
    >>Thanks a lot,
    >>Federico
    >
    >
    > Hmmm .... assuming that you mean at the server end. If you mean one
    > process per client, just use/require cookies.
    >
    > Remember that http is stateless -- you deliver a page that indicates
    > it can launch a program, but back at the server it's a no-go :-(
    >
    > How do you let the client know what is going on and how do let them
    > know it's OK? You can't!
    >
    > Now, if the program is small enough that it runs to completion quickly
    > (like a typical cgi request), thus opening the possibility for the
    > next queued request to run, you _may_ get by...
    >
    > But what if the program hangs, returns an error, fails to release
    > resources indicating it's busy, etc.? What if the client leaves while
    > waiting?
    >
    > I think I would look for a better solution before going down this
    > road. Preferably you would handle all contingencies on the server
    > without having to truck back and forth informing an impatient client
    > why it's not "working" as they expect. Cryptic error messages do not
    > confidence build ...
    >
    > If you were more specific with what you're trying for and why you
    > think this is the best/only way to go (eg., the "program" to run is
    > out of your control), then I would feel more confident offering
    > suggestions. As is, I'm not sure what to offer other than that I
    > don't care for this straight, linear (serializing/fifo) of requests
    > (even with semaphores and polling) -- it can and _is_ often done, but
    > requires more work than you might imagine. It's why _concurrency_ is
    > such a big deal with databases (and difficult to implement) in a way
    > that keeps everbody happy.
    >
    > Why can't you spawn/fork a different process for each request?
    > Usaully this is a multiple writers problem -- is this the case?
    >
    > If it's that important, why not a single process that launches mutiple
    > threads? Agreed, these are not easy to carry off, but your idea (over
    > the web) might even be _more_ trouble to implement. Clients don't
    > like to hear, "wish I could help, but stuff has happened at this end
    > that means you have to wait or try again later".
    >
    > I know this doesn't help in a concrete sense, but, like I said, it's
    > hard to know how/where to go from what you've said.
    >
    > regards,
    > prg
    > email above disabled

  • Next message: Federico: "Re: queue system"

    Relevant Pages

    • Re: breaking the model
      ... > The forms data then is in the Request object. ... HTTP Request; in this case, the form POST Request from the Page. ... client and server. ...
      (microsoft.public.dotnet.framework.aspnet)
    • Re: Resolving record with enumerated type
      ... In a CPU BFM package, ... because data goes in two directions (request from the ... from the server to the client), you'll need some way to orchestrate ...
      (comp.lang.vhdl)
    • Re: NadaNet file server coming soon...
      ... Applesoft BASIC running under ProDOS, in less than 300 lines, ... After each command is completed by the server, ... client the result code, the AUX TYPE, ... forgoes queueing (since one client can't have more than one request ...
      (comp.sys.apple2)
    • Re: Kerberos with Windows Integrated authentication
      ... behaviour if your Web server is in the client broweser's Internet zone. ... referencing it by computer name rather than FQDN), the browser will request ... Obviously, if you want to use Kerberos for authentication, you will either ...
      (microsoft.public.windows.server.security)
    • Re: Performance Issue with Runtime Image
      ... >> the client, closes the connection, then dies. ... request before even accepting the next incoming connection. ... The client program is unaffected so presumably the server is ...
      (comp.lang.smalltalk.dolphin)