A simple virtual organisation model and practical implementation - simulation visualised

2017-02-24T11:42:07Z (GMT) by Lyle Winton
The VO managed job queueing service prototype was<br>tested in a simulation together with the previously<br>mentioned resource level job consumer. The simulation<br>included 3 VOs each with 10 users of varying VO<br>priority. The VOs maintained one queueing service<br>each. All VOs have access to 10 simulated resources,<br>each consisting of a clusters of 10 to 50 nodes or job<br>execution slots. Each resource allocated a different<br>local priority range for each VO queue. The ranges<br>were of at least 10 and randomly allocated between<br>1 and 100. Each user periodically submitted between<br>10 and 50 jobs at once but only when they had no<br>jobs in the queue. Each job took between 1 and 10<br>minutes to complete on any node.<br>This particular scenario led to an average stable<br>state of 30 jobs for each of 30 users. These were shared<br>across 10 hosts with an average of 30 nodes. The<br>average job load for this situation is 3 times more<br>than the number of available nodes, an over utilised<br>resource scenario. Resource usage reached saturation<br>within 4 minutes of the simulation commencing. The<br>state of the queue was frozen after 15 minutes for<br>evaluation.