IPM logo
Reza Entezari-Maleki, Ph.D.
 

Postdoctoral Researcher

School of Computer Science

Institute for Research in Fundamental Sciences (IPM)

 
 
 


Research Interests
  • Performance Modeling and Evaluation: Performance is the key issue for any system that services user requests. Modeling the performance of a computer system, system providers can analyze the current system and apply modifications to improve the performance if it is required. Performance modeling and evaluation can also be considered as a design tool, when the system under study does not still exist. There are several mathematical approaches, used as modeling tools, to evaluate the performance of a computer system. Our main tools for modeling the performance are stochastic extensions of Petri nets. Applying Stochastic Reward Nets (SRNs) and Stochastic Activity Networks (SANs), we can appropriately model and analyze the performance measures of various computing systems. In some cases, Markov Chains (MCs), Markov Reward Models (MRMs), and Queueing systems help us to achieve our goal.

  • Dependability and Performability: Dependability of a system is the ability to avoid service failures that are more frequent and more severe than is acceptable. Actually, dependability is an integrating concept that encompasses measures availability, reliability, safety, integrity, and maintainability. Dependability analysis, especially in the case of computer systems and networks, is an interesting research area which has attracted many researchers. In addition to modeling and evaluation of dependability measures (e.g. reliability and availability), we are interested in simultaneously studying the performance of a system. Combined performance and dependability analysis of fault tolerant computing systems which is known as performability analysis is a more interesting issue in modern large-scale distributed systems in which their resources are prone to failure. We use mathematical models to simultaneously analyze the performance and dependability of computer systems.

  • Grid and Cloud Computing: Grid computing is a technology to build dynamically constructed proble solving environments using geographically and organizationally distributed computational resources connected via communication links. Grid computing provides supercomputing like power on demand, just as a power grid which provides electricity on demand. Based on definition, a computational grid is a large collection of computers (computing resources) linked via the Internet (any network) so that their combined processing power can be harnessed to work on difficult or time consuming problems. The term cloud computing came into popularity which has some conceptual similarities with the grid computing. There are some problems in grid and cloud computing we are interested in studying them. For example, performance/dependability evaluation of grid manager and distributed resources in grid, efficient task scheduling among grid resources, modeling the virtual machine migration in clouds, studying the virtualization and rejuvenation in cloud computing, analyzing the Web service composition in clouds, assessing power consumption, and so forth.

  • Task Scheduling Algorithms: Task scheduling is one of the well-known problems in computer systems involving dispatching incoming requests to the resources with the aim of achieving a predefined goal. We are interested in this problem when it is applied to the distributed computing systems. Since computing resources in a large-scale distributed system are heterogeneous and locally distributed, task scheduling plays an important role in achieving high performance/throughput computing. There have been proposed many static and dynamic scheduling algorithms to schedule tasks among the resources to achieve the required amount of Quality of Service (QoS). Our main focus is on proposing novel and state-of-the-art scheduling algorithms and disciplines to increase the performance of the system and achieve satisfiable QoS. In some cases, meta-heuristics are used to attain the goal since the problem is NP-complete.

 

 

 

 

 

 

 

 

 

 

 


Update