Hi Terry,

I might have solved your problem for you already.

Try taking a look at: http://github.com/TheLastCylon/kisscpp

Documentations, such as it is, is available here: http://thelastcylon.github.io/kisscpp/index.html

Currently, there is no UDP support, but I'm sure it won't take much to get that in.
Feel free to ask questions, if you want.

--
Regards,
Dirk J. Botha
http://www.djb.co.za
~*~ Registered Linux User #379726 ~*~

"If people concentrated on the really important things of life, there'd be a shortage of fishing poles." ~ Doug Larson


On 30 December 2013 08:41, U.Mutlu <for-gmane@mutluit.com> wrote:
Terry Lazoki wrote, On 12/29/2013 03:43 AM:

I'm having hard time understanding the boost asio design and how i can use
it to tackle my problem.

My problem:
Designing a tcp/udp server that is time efficient.

1. User sends packet through tcp
2. Server receives packet and processes it to build a large tree (n > 20000)
3. Server then parses the new tree and sends a summary packet through udp
to another client

Conditions/Limitations:
  - Has to be time efficient, receiving, processing and sending has to be
done as fast as possible.
  - Processing the packet is heavy and can cause delays if its done in the
same thread
  - Parsing the tree is also a little heavy when sending it through udp, can
cause delays too

Design 1 (2 threads):
  Thread 1: ASIO TCP and UDP. TCP receives packets and adds it to a circular
buffer. While at the same time, processes the shared tree and sends packet
through udp.
  Thread 2: Server grabs packet from circular buffer and processes it

Some problems here:
  - I dont understand the async part here. What happens when tcp receives a
packet, if the current thread is continuously parsing the tree and sending
packets through udp. Do i need to separate the udp part to another thread?
  - I have 3 parts of the program that i need to run concurrently, well its
really 2 parts and receiving packets through tcp.


Any help is appreciated.

Thanks,


I would use a modular approach using an acceptor, a shared job queue,
and multiple worker threads (at least 2 workers; depends on the job
characteristics, duration etc.). Of course a synchronisation/locking
is needed for shared resource access/usage,
see for example https://en.wikipedia.org/wiki/Producer–consumer_problem

--
U.Mutlu
DACOS Notdienstanlagen GmbH, Germany, www.dacos.de



_______________________________________________
Boost-users mailing list
Boost-users@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users