Active XML

AXML Continuous Services Benchmarks

Computer Specs

  • Computer 0,1 Specs:
    Intel Pentium 4 3.0 GHz
    512 MB RAM
    sun jdk 1.5.0_03
    tomcat 5.0.27
    Linux Caixa Mgica Servidor 10
  • Computer 2 Specs:
    Intel Pentium 4 2.66 GHz
    512 MB RAM
    sun jdk 1.5.0_03
    tomcat 5.0.27
    Linux Caixa Mgica Servidor 10

Description

Computer 0

  • Runs a axml peer application deployed on a tomcat server
  • Runs a virtual machine with a axml peer application in a tomcat server.

Computer 1,2

  • Runs a script that invokes a continuous call, the callback is redirected to the peer running in the virtual machine on Computer 0.

Benchmarcks

Case 1 100 sequential invocations of the continuous call
Case 2 100+100 concurrent invocations of the continuous call
Case 3 100*4 concurrent invocations of the continuous call
Case 4 100*8 concurrent invocations of the continuous call

The 4 above test cases are repeated in parallel on both computers 1,2 , i.e,

Benchmark the response time to invocations from one computer

    Computer      |    Computer 0             |
                  |                           |        
    case 1 -----> |                           |
                  | tomcat manually restarted |
    case 2 -----> |                           |
                  | tomcat manually restarted |
    case 3 -----> |                           |
                  | tomcat manually restarted |
    case 4 -----> |                           |

Benchmark the response time to invocations from two computers in parallel

    Computer      |         Computer 0        |    Computer 2
	          |                           |        
    case 1 -----> |                           |  <----- case 1 
	          | tomcat manually restarted |
    case 2 -----> |                           |  <----- case 2 
	          | tomcat manually restarted |
    case 3 -----> |                           |  <----- case 3
                  | tomcat manually restarted |
    case 4 -----> |                           |  <----- case 4

All the above cases have been repeated twice.

Results

Description of the result files

The result files have two section with the benchmarks made with one computer and the benchmarks made from two computers. All results are in seconds.

================ 1 computer =================

case 1 
 average 0.0461583590508 (average time for 100 invocations)
 
case 2 
 average 0.0792604398727 (average time for 100 invocations)
 average 0.0809775400162 (average time for 100 invocations)
------------------------
 AVG: 0.080118989944449992 (average time for all (200) invocations)
    
case 3 
 average 0.107943139076
 average 0.109124617577
 FAULT
 average 0.104744906425
-----------------------
 AVG: 0.10727088769266667
	 
case 4 
 average 0.192550077438
 average 0.188536131382
 average 0.184168522358
 average 0.174193813801
 average 0.185034630299
 average 0.168694000244
 average 0.180873212814
 FAULT                    (fault while creating a SOAP Proxy)
-----------------------
 AVG: 0.18200719833371429
		  
  1 Computer 2 Computers
Total AVG Total AVG
1 client (100 requests) 4.61 0.0459799051285 7.23 0.056175028085749992
2 clients (2*100 requests) 8.53 0.0844246304035499 11.15 0.103685448765775
4 clients (4*100 requests) 12.31 0.12311121523375 17.83 0.17432741999599999
8 clients (8*100 requests) 19.42 0.19305380702012503 29.55 0.28205116625343751*
  • 1 fault occurred while creating a SOAP proxy, 700 requests made from the computer 2.

More details about the first benchmark here

  1 Computer 2 Computers
Total AVG Total AVG
1 client (100 requests) 4.73 0.0461583590508 6.63 0.057109016180050001
2 clients (2*100 requests) 8.21 0.080118989944449992 11.34 0.10725002884849999
4 clients (4*100 requests) 10.97 0.104744906425 * 18.51 0.17904788792124998
8 clients (8*100 requests) 19.12 0.18200719833371429 27.56 0.26804782550256251
  • 1 fault occurred while creating a SOAP proxy, 300 requests made from the computer 1.

More details about the second benchmark here


Original document from Caixa Magica is here.


For any questions about this page, please contact Gabriel Vasile.