"Stig S. Bakken" wrote: > > Edin Kadribasic wrote: > > > > On Wed, 5 Dec 2001 [EMAIL PROTECTED] wrote: > > > > > On Wed, 5 Dec 2001, Edin Kadribasic wrote: > > > > > > > The trouble with remote object access mechanisms is that most of them are > > > > really slow. I had great success with serializing objects and arrays into > > > > shared memory. All of that was done in PHP using sysvshm extension. Maybe > > > > some system support for that could make it even easier and hopefully > > > > even faster. > > > > > > The 'remote' accessesible objects can also be on the same server, with > > > communications trhough unix domain sockets, this shouldn't be that slow. > > > > Well, I'm aware of that :) Still, you can serialize objects to the > > database if you want to, but the performance is going to be poor. I cannot > > image XMLRPC being any faster. I did use Corba a while back, and I'm still > > recovering from the experience. > > I am not claiming that neither XML-RPC, SOAP or Corba is fast, but > rather ways of solving this type of problem. > > > > > The approach is not perfect. Your objects become readable to every > > > > user on the same webserver, so it would probably be impractical for ISPs > > > > to enable it. > > > > > > In SRM objects are bound to a session (if that is requested from the > > > script while creating an object). > > > > I have not tried that before. To be honest, I'm clueless as to what SRM > > is. Could you please send some pointers? > > http://www.vl-srm.net/ > > I think SRM will be able to deliver acceptable speed. Also, Mark > Woodward's msession extension is developing in the same direction, and > msession already has proven itself very fast (1500+ requests per second > on cheap intel hardware).
Stig have you tested that number? I'm curious to know if others are getting the same results as me. I can get those numbers as long as the data fits in one or two MTU, obviosly if you are sending a much more data it will be slower. Also, the performance I got hovered around 1400 connections a second. The msession system was written BECAUSE databases and solutions like Corba were slow. I neded an equivilent of shared memory. While not that fast, it is pretty fast. (msession is actually the rebirth of some code I had been thinking about for a homogenous beowulf cluster.) The latest version on my website supports a simple plugin system. The system in development supports a much more flexable plugin system which allows end user code to be called. -- PHP Development Mailing List <http://www.php.net/> To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] To contact the list administrators, e-mail: [EMAIL PROTECTED]
