Hp loadrunner test on this website
	The website you are looking at is running on C++ code. The code was made to be easy and fast deployable on any Windows machine. The code should meet the following demands.
	• Adjustable resource usage
	• The web server should perform using less the 12 Mbytes.
	• Transaction response time should be under 1 second when servicing 5 simultaneous users.
	• The program should have no memory leakages
	• The program should have an uptime of more the one month.
	I am going to test this web server (1.0.55.1) to see whether the code meets these demands. The code will be tested with a connection test to see whether connections are build up fast enough and to see whether any sockets are left open. This will prove that there is no degrading of the web server to the aspect of sockets. The uptime of one month will be tested with a stress test of one day doing 30 times the normal day load. The web server will be restarted by another service in case of a crash.
	The connection test will have the following scenario.
	The connection test should be done with a very small request and response to ensure no interference of the downloading and processing of the data. I know that the web server is returning its time in a small response as a result of a small request. The request and the response look like this.
	Request
	GET /DateTimeT045R000000006.js HTTP/1.1
	Referer: http://www.stefanboersen.nl/Contact/PerformanceTest.html
	User-Agent: Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 6.0; SLCC1; .NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 3.5.21022)
	Accept-Encoding: gzip, deflate
	Accept-Language: en-us
	Connection: Keep-Alive
	Host: www.stefanboersen.nl
	Response
	HTTP/1.1 200 OK
	Content-Type: application/x-javascript
	Content-Length: 82
	Connection: keep-alive
	ZZAgentWebThreadId: 23
	function FT045R000000006()
	{
	document.forms[0].nameTime2.value = 35495.297;
	}
	Script used for the test
	Action()
	{
	double atof (const char *string); /* Explicit declaration */
	web_set_max_html_param_len("1024");
	web_reg_save_param("paramThreadId","LB=ZZAgentWebThreadId: ", "RB=\r\n", LAST);
	web_reg_save_param("paramRequestId","LB=R", "RB=()", "Ord=1", "Search=body", "RelFrameId=1", LAST);
	lr_start_transaction("01_SB1_connect");
	web_url("DateTimeJs",
	"URL=http://www.stefanboersen.nl/DateTimeT045R000000006.js",
	"Resource=1",
	"RecContentType=application/x-javascript",
	"Referer=http://www.stefanboersen.nl/Contact/PerformanceTest.html",
	"Snapshot=t49.inf",
	LAST);
	lr_end_transaction("01_SB1_connect",LR_AUTO);
	lr_start_transaction("02_SB1_wait");
	lr_think_time(1);
	lr_end_transaction("02_SB1_wait",LR_AUTO);
	lr_user_data_point("paramThreadId",atof(lr_eval_string("{paramThreadId}") ) );
	lr_user_data_point("paramRequestId",atof(lr_eval_string("{paramRequestId}") ) );
	return 0;
	}
	Result of the test
	Discussing the result of the test
	The last picture is a very well know result. The blue lines , indicated by the red arrows 1,2,3, explain a lot. During line 1 the number of virtual users is increased and the response time stays the same. During line 2 the number of virtual users is increased and the response time is increased proportionally. During line 3 no more virtual users are added and the response times are not increasing. If there is degrading of the system you would see increasing response time although no more virtual users are added. At point 4 the number of virtual users is increased and the response time stays the same. 5 is the name of the transaction and 6 is the average connection time.
	The test is doing a throughput 450 bytes per second. This is not very much , we are doing a connection test. The test did 15549 connections in two hours. This is not very realistic. Any user of the site would connect and then would start reading.
	Any new connection is handled by another thread. This ensures that there was a new connect. The setting to close instead of keep-alive is easily forgotten. The property to be set is here:
	What would be the result of the test if we had forgotten the close setting.
	Increasing the number of virtual users would not influence the transaction response time. The test asked the server the time and did that 36169 times in two hours.
	The requests are all handled by the same threads.
	stefanboersen@hotmail.com