Re: [Mono-dev] Current Implementation of Async Sockets
Would there be any mileage to looking at libuv? It seems to wrap IOCP and libev. Its not clear whether the limited performance implied here (http://www.yesodweb.com/blog/2011/03/preliminary-warp-cross-language-benchmarks) is a reflection of the single-threaded node engine, or more intrinsic. I'd expect the former; libev is designed for reactive use in a single thread environment, but does allow you to run multiple event handlers so you should be able to run a handler per core. It would be nice if mono aio could approach the performance of Java nio, but i don't think it ever has. ___ Mono-devel-list mailing list Mono-devel-list@lists.ximian.com http://lists.ximian.com/mailman/listinfo/mono-devel-list
Re: [Mono-dev] Current Implementation of Async Sockets
Oh, this looks like a rather advanced one! After some scanning through the sources he seems to have written the server in c# on top of mono and then he extended (or modified) the implementation of monos socket implementation. It's not quite the type of server I want to write and I might be a little overwhelmed by what was done there, but I'll have a look. I would prefer thought, if the implementation in mono worked performant. I found some links which hint on some recent? changes to that implementation, like that one: http://c-cpp.r3dcode.com/files/mono/2/10.2/mono/metadata/tpool-epoll.c It's really hard to find updated documentation on the project. Am I looking in the wrong places? On the website, most of the stuff is seriously outdated. -- View this message in context: http://mono.1490590.n4.nabble.com/Current-Implementation-of-Async-Sockets-tp4642031p4644669.html Sent from the Mono - Dev mailing list archive at Nabble.com. ___ Mono-devel-list mailing list Mono-devel-list@lists.ximian.com http://lists.ximian.com/mailman/listinfo/mono-devel-list
[Mono-dev] Current Implementation of Async Sockets
Hi, I am currently doing some research on the best platform for a high performance socket server. I want it to be capable of handling a large number of small requests and a high number of Clients at the same time. So far I've read, that I/O Completion Ports are implemented pretty well under Windows but Linux seems to use another model. In some other thread I found, that implementation was a rather badly performing wrapper in Mono under Linux. Has this changed in the last time? Is it good to use Mono for that purpose or should I rather stick with Java? I'm also thinking of going the C++ way, but I'm not yet sure if the result would justify the additional work. Thank you, Marius -- View this message in context: http://mono.1490590.n4.nabble.com/Current-Implementation-of-Async-Sockets-tp4642031.html Sent from the Mono - Dev mailing list archive at Nabble.com. ___ Mono-devel-list mailing list Mono-devel-list@lists.ximian.com http://lists.ximian.com/mailman/listinfo/mono-devel-list
Re: [Mono-dev] Current Implementation of Async Sockets
On 16 May 2012 16:33, longway77 longwa...@gmail.com wrote: Hi, I am currently doing some research on the best platform for a high performance socket server. I want it to be capable of handling a large number of small requests and a high number of Clients at the same time. So far I've read, that I/O Completion Ports are implemented pretty well under Windows but Linux seems to use another model. In some other thread I found, that implementation was a rather badly performing wrapper in Mono under Linux. Has this changed in the last time? Is it good to use Mono for that purpose or should I rather stick with Java? I'm also thinking of going the C++ way, but I'm not yet sure if the result would justify the additional work. It's possible, see https://github.com/jacksonh/manos -- Michael Hutchinson http://mjhutchinson.com ___ Mono-devel-list mailing list Mono-devel-list@lists.ximian.com http://lists.ximian.com/mailman/listinfo/mono-devel-list
