Hi we are using kannel about 2 years in our institution and its woriking like a charm. we have high load traffic with 3 operators with 3 kannel gatways, our configuration is like this:
- kannel 1 : Operator 1 : VM 2 x CPU 3 Ghz 8 Go Ram - kannel 2 : Operator 2 : VM 2 x CPU 3 Ghz 8 Go Ram - kannel 3 : Operator 3 : VM 2 x CPU 3 Ghz 8 Go Ram - Web Portal: for all kannel : VM 8 x CPU 3 Ghz 32 Go RAM - Postgres Database behind the web server : VM 8 x CPU 3 Ghz 32 Go RAM All those server are running under ESXi hypervisor in SAN environnement. Any way we are tunning all the plateform with several tools ( ApacheBench for web load and fakesms for kannel) we want to profit of all the hardware ressources with optimizing all componenents to reach the limits of the hardware, but still are still far behind the real capability of the hardware. The fact is we have reached 800 sms /s with the operators and its fine, but we want more not that we need it right now but just to master how every things work... The WEB/DB side tunning is "well" done due to fact that we have "mastered" it in our webs servers but the kannel side is litle different. with fakesms we are sending about 1000 sms / s from each kannel so we want to manage 3000 sms / s in our portal but all the kannel dosent send them with the speed we want even if they are not even litle solicited, we noticed that the bottle neck is between the smsbox and the bearerbox we dont know exactly but all 1000 sms arrive instantelly to the bearerbox in each kannel and a queue is formed between the bearerbox and smsbox, the smsbox send to the web server ( the portal ) about 1200 request / s wich is far from his limit ( we have tunning it to manage 3500 request / s ) I ve talked a lot but the main question is how we remove that so called queue smsbox/bearebox? to make the kannels send more than 1200 request/s so we can reach the limits in all the equippements all the system is running far behind his capabilitys. thinx for the help
