check out
http://grizzly-sendfile.kenai.com/
it does what you want for downloads (few worker threads + ton of
concurrent client + large files).
let me know if that works and I'd be happy if you decided to
contribute to the project if you see something missing.
/i
On Thu, Aug 19, 2010 at 6:37 AM, Vrignaud Etienne <evrignaud_at_axway.com> wrote:
>
> Hi folks,
>
> I am using Grizzly 1.9.19-beta7.
> My use case is that the clients of my application needs to upload or
> download very big files (more than 50 Go).
> Every time a client do an upload or a download of that kind of big file, one
> thread of the Grizzly http server is busy during many times, and only
> dedicated to this task.
> The consequence is that, if many clients are doing upload or download at the
> same time, on the server part all the Grizzly threads dedicated to manage
> requests are busy in receiving or sending the files. And the server does not
> answer any more to the simple http client requests. They have to wait until
> some upload or download are finished.
>
> I have seen that there is by default 5 threads that manage http requests.
> Using selectorThread.setMaxThreads() I succeed to increase this value to
> 256.
> Unfortunately the 256 value does not seem to fit my needs, because it
> consumes too many resources on my server (mainly memory, TCP ports), and
> also I can't imagine to limit the number of simultaneous uploads/downloads
> to 256.
> I cannot set this kind of limit in my application.
>
> Could you tell me what are the best ways to manage this kind of edge case
> with Grizzly.
> Are there some dedicated thread pools that can manage file upload/download
> in parallel?
> Are there some internal mechanisms in Grizzly that could be used to enable
> NonBlocking IO to receive or send the files?
>
> At this time my implementation for the upload is based on
> commons-fileupload, and for download I simply return an OutputStream as
> request response.
>
> My big requirement is that on the client part, a browser must be enough to
> download the file. This implies that it's not possible to use my own
> algorithm to split the file in blocks.
>
> Thanks for your help,
> /Etienne VRIGNAUD
>
> --
> View this message in context: http://old.nabble.com/Grizzly-no-more-responding-when-clients-are-doing-many-big-files-upload-and-download-tp29482040p29482040.html
> Sent from the Grizzly - Users mailing list archive at Nabble.com.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>
>