Re: [xSocket-develop] scalability and latency

[Gregor R. <gre...@go...>]

Hi James,

this is a typical scenario which should be supported by xSocket-http.

Unfortunately, performance (scalability, latency, …) diagnostics &
optimizations haven't been done for xSocket-http yet (for resource/time
reasons). I would suspect, that the new module xSocket-http causes this
undesired effects.

I have to take a deeper look into this. Could you send me minimal code, that
I'm able to run some test and make some diagnostics?

Which platform do you use (windows, linux,…)?

Gregor

2008/4/4, James Olsen <Jam...@co...>:
>
>  I'm experimenting with xSocket 2.0-beta-2 and xSocket-http 2.0-alpha-4 to
> implement a messaging server that pushes messages to http clients. The
> messages are small (approx 15 bytes) but frequent (enqueued at a rate of 25
> per second per client).  These messages are in 10 logical topics and only
> the latest available message from each topic is sent, i.e. when a new one is
> queued to be sent it overwrites any message already queued for the same
> topic.  The server infrastructure will also concatenate several smaller
> messages to make one larger message if messages are available on multiple
> topics at the time of writing to the BodyDataSink.  Thus the number of
> messages dequeued and sent to the clients is automatically throttled by the
> capability of the underlying comms.
>
>
>
> I have the same setup with Tomcat 6 using the NIO connector and find that
> I can add large numbers of users and still get good performance.  With the
> hardware I have it can deliver 70,000 messages per second.  More users mean
> fewer messages per user but the overall throughput remains the same and
> latency is low and fairly consistent.
>
>
>
> However I'm finding that this does not scale as well with xSocket.  The
> max message throughput is lower and reduces as more clients are added.  The
> average latency is higher and can be highly variable (>10 seconds!)
> depending on the number of background message writing Threads and the number
> of Dispatchers.  Making the number of message writing Threads equal to the
> number of Dispatchers and keeping them both quite low (<= 3) reduces the
> latency variability but the average is still significantly higher than with
> Tomcat.  SYNC flushing is also better than ASYNC with this setup because it
> causes more message concatenation (fewer larger messages sent).
>
>
>
> The xSocket API has been a joy to work with and allows me to do more than
> with Tomcat/Comet (e.g. if desired the client can continue to write to the
> connection) and the memory use is incredibly low, however the performance
> for my scenario is worse than I was expecting.
>
>
>
> Is there something I might be missing in my config or is this not a
> scenario that xSocket is currently optimal for?
>
>
> ***********************************************************************************************
> This email, its contents and any files attached are a confidential
> communication and are
> intended only for the named addressees indicated in the message.
>
> If you are not the named addressee or if you have received this email in
> error, you may not,
> without the consent of Cognotec, copy, use or rely on any information or
> attachments in any way.
> Please notify the sender by return email and delete it from your email
> system.
>
> Unless separately agreed, Cognotec does not accept any responsibility for
> the accuracy or
> completeness of the contents of this email or its attachments. Please note
> that any views,
> opinion or advice contained in this communication are those of the sending
> individual
> and not those of Cognotec and Cognotec shall have no liability whatsoever
> in relation to
> this communication (or its content) unless separately agreed.
>
> ***********************************************************************************************
>
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
>
> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
> _______________________________________________
> xSocket-develop mailing list
> xSo...@li...
> https://lists.sourceforge.net/lists/listinfo/xsocket-develop
>
>