Abstract:
On-demand computing is a popular enterprise model in which the computing resources are made
available to the users as needed. On-demand computing based transaction processing system which
has grown rapidly in recent years is an information processing system with the stringent requirements
of resources to meet the fluctuating demands. Concepts such as grid computing, utility computing,
autonomic computing, and adaptive management seem very similar to the concept of on-demand
computing. When demands of resources fluctuate, the system needs load balancing for the efficient
utilization of the computational resources. Furthermore, scheduling is needed to assign the transactions
to the appropriate resources. Thus, modeling of load balanced scheduling along with reliability analysis
for this system is a challenging task.
This paper presents the load balanced scheduling and reliability modeling in such an environment by
using colored Petri nets (CPNs). CPNs which combine Petri nets with programming languages is a powerful
modeling technique. The proposed CPN-based modeling pattern formally describes the process of transaction distribution and execution within the on-demand computing environment. Moreover, the CPNbased model uses the hierarchical modeling capability of CPNs, including different levels of abstraction
(sub-modules). This helps easily handling and extending the model. Since, on-demand computing based
transaction processing system executes a number of concurrent transactions. The CPN-based model is
extended to express the concurrency, thus improving the reliability results. This paper takes the example
of grid transaction processing (GTP) system with the problem of load balanced scheduling modeling and
reliability evaluation.
