Performance Evaluation of a SNAP-Based Grid Resource Broker

Deployment and Performance Evaluation of a SNAP-based Resource Broker on the White Rose Grid

“…In this architecture, the broker begins by parsing the user requirements submitted through a Grid portal; this is the first layer of SNAP Task Service Level Agreement (TSLA) [4]. The second layer, Resource Service Level Agreement (RSLA), uses a Matchmaker, supplied with the parsed user requirements, to contact a Knowledge Bank (KB).…”

“…In [4], the problem of oscillation that could occur between the broker and the information provider without a successful job submission was highlighted. The key issue outlined is that information obtained from the information provider for each resource may be out-of-date by the time a decision is made as to where the job should run and the Co-ordinator proceeds to attempt to reserve the chosen resources.…”

“…To support application execution in the context of the Grid, it is desirable to have a resource brokering service [9]. In [4] the authors discussed a simple SNAP (Service Negotiation and Acquisition Protocol) based resource broker and a more sophisticated SNAP broker, following a threephase commit protocol. SNAP [13] is an appropriate choice in the design and implementation of a usercentric broker, since it provides the means to negotiate and acquire resources that meet the user's application requirements through Service Level Agreements (SLA).…”

“…Experimental results obtained through mathematical modelling and simulation have shown that the use of the three-phase commit SNAP broker provides a performance enhancement over a simple SNAP broker, in terms of the time it takes for a job to be successfully submitted and begin execution [4]. However the issue of how well the three-phase commit SNAP broker compared to the simple version performs over a large distributed Grid infrastructure has not been addressed and is the subject of this paper.…”

“…An example of a large infrastructure is the White Rose Grid (WRG) which consists of high performance computing resources at Leeds, Sheffield and York Universities [14]. The WRG exhibits heterogeneous resources and spans over multiple administrative domains, providing an environment which differs from previous studies [4] due to its true nature as a Grid infrastructure. Experiments are carried out and performance results show that the inclusion of the three-phase commit protocol provides a performance enhancement in terms of the time taken from submission of user requirements until a job begins execution.…”