Load Test with RTC Simple JSON Server and RTC Stress Tool v 2.0 from Demos

[RMC]

Hello Danijel

i have this scenario:

Server : dedicated Server

Intel Xeon(R) CPU E5-2620 2 GHz Quad Core
RAM 24 GB
Windows 2008 R2 Server
64 bit OS

Simple RTC Json Server started with:
Multithreaded... checked
MaxThreads 50
Generate Json Results... checked
Generate custom HTTP Headers... checked

Clients PC's: 5 VServer where i have started several RTC StressTools v.20 each with 20000 Requests per loop, totally 2021 active connections

MaxThreads: 50
Use AutoConnect ... checked
Disconnect after loop ... checked
AutoRepeatLoops ... checked
AutoConn+ : 1

------------------------

Now when i want to start from the 6th VServer another RTCStressTool i get new connections but all with Requests Errors.

------------------------

Please tell me how can i find the problem, why i get this Requests Errors? Those 2021 active connections are the limit or i must do some another settings?

Thank you
Radu

[D.Tkalcec (RTC)]

Server needs a free thread to do something. When all the threads are busy working, as is the case in stress-test scenarios, the Server will stop accepting new connections as long as there is a lot of load on the Server. How many active connections and requests the Server will be handling at the same time will depend on a lot of factors, like the implementation of your Server, the number of threads you allow in your thread pool and the CPU power available on the Server. Also, you need to know that there is a limit on the number of TCP/IP ports on the Server and each TCP/IP port used by a Client will remain occupied on the Server for some time after the connection is closed. What you are doing with your test is basically a DDOS attack on the Server, where a number of Clients are opening and closing a lot of connections and sending bursts of requersts. The errors you are getting on the Clients are either caused by your network being overflooded and unable to handle more load, or connections which get dropped because the Server does not have enough power to process all the requests on time, or because your Clients are overburdened.

Anyway ... is your Scenario something that is going to happen in your real-life application, or are you just testing the limits on the RTC SDK?

Best Regards,
Danijel Tkalcec

[RMC]

Thank you Danijel.

I followed your advices and now i have 30 000 active connections on the server .

Radu

[RMC]

Hi Danijel

Sorry because just i answer on your question, i test this Scenario for a real life application but of course with this scenario i test also the limits of the RTC SDK.

best regards
Radu

[D.Tkalcec (RTC)]

Even though the RTC SDK can technically handle thousands of active connections, there are limits of the TCP/IP protocol (like the number of available ports), memory llimits and memory management issues as memory usage goes up (performance drops), then there are CPU limits, network usage limits, database limits and other issues which you will most likely encounter if your Server has to handle thousands of active Clients. When you get to the point that a single Server can not handle the load from all your Clients and there is nothing else you can do to make a single Server better, I recommend you to take a closer look at the Load Balancer and distribute the load on multiple Servers.

Best Regards,
Danijel Tkalcec

[RMC]

Hi Danijel

thank you, i intend to use also the LoadBalancer for more servers.

best regards
Radu