List of Articles Load Balancing

• Open Access Article
  • Abstract Page
  • Full-Text
  
  1 - Network RAM Based Process Migration for HPC Clusters
  Hamid Sharifian msharifi msharifi
  10.7508/jist.2013.01.006
  Process migration is critical to dynamic balancing of workloads on cluster nodes in any high performance computing cluster to achieve high overall throughput and performance. Most existing process migration mechanisms are however unsuccessful in achieving this goal pr More

  Process migration is critical to dynamic balancing of workloads on cluster nodes in any high performance computing cluster to achieve high overall throughput and performance. Most existing process migration mechanisms are however unsuccessful in achieving this goal proper because they either allow once-only migration of processes or have complex implementations of address space transfer that degrade process migration performance. We propose a new process migration mechanism for HPC clusters that allows multiple migrations of each process by using the network RAM feature of clusters to transfer the address spaces of processes upon their multiple migrations. We show experimentally that the superiority of our proposed mechanism in attaining higher performance compared to existing comparable mechanisms is due to effective management of residual data dependencies. Manuscript profile
• Open Access Article
  • Abstract Page
  • Full-Text
  
  2 - Load Balanced Spanning Tree in Metro Ethernet Networks
  Ghasem Mirjalily Samira Samadi
  10.7508/jist.2014.02.007
  Spanning Tree Protocol (STP) is a link management standard that provides loop free paths in Ethernet networks. Deploying STP in metro area networks is inadequate because it does not meet the requirements of these networks. STP blocks redundant links, causing the risk of More

  Spanning Tree Protocol (STP) is a link management standard that provides loop free paths in Ethernet networks. Deploying STP in metro area networks is inadequate because it does not meet the requirements of these networks. STP blocks redundant links, causing the risk of congestion close to the root. As a result, STP provides poor support for load balancing in metro Ethernet networks. A solution for this problem is using multi-criteria spanning tree by considering criterions related to load balancing over links and switches. In our previous work, an algorithm named Best Spanning Tree (BST) is proposed to find the best spanning tree in a metro Ethernet network. BST is based on the computation of total cost for each possible spanning tree; therefore, it is very time consuming especially when the network is large. In this paper, two heuristic algorithms named Load Balanced Spanning Tree (LBST) and Modified LBST (MLBST) will be proposed to find the near-optimal balanced spanning tree in metro Ethernet networks. The computational complexity of the proposed algorithms is much less than BST algorithm. Furthermore, simulation results show that the spanning tree obtained by proposed algorithms is the same or similar to the spanning tree obtained by BST algorithm. Manuscript profile
• Open Access Article
  • Abstract Page
  • Full-Text
  
  3 - A Novel Approach for Cluster Self-Optimization Using Big Data Analytics
  Abbas Mirzaei Amir Rahimi
  10.7508/jist.2019.01.005
  One of the current challenges in providing high bitrate services in next generation mobile networks is limitation of available resources. The goal of proposing a self-optimization model is to maximize the network efficiency and increase the quality of services provided More

  One of the current challenges in providing high bitrate services in next generation mobile networks is limitation of available resources. The goal of proposing a self-optimization model is to maximize the network efficiency and increase the quality of services provided to femto-cell users, considering the limited resources in radio access networks. The basis for our proposed scheme is to introduce a self-optimization model based on neighbouring relations. Using this model, we can create the possibility of controlling resources and neighbouring parameters without the need of human manipulation and only based on the network’s intelligence. To increase the model efficiency, we applied the big data technique for analyzing data and increasing the accuracy of the decision-making process in a way that on the uplink, the sent data by users is to be analyzed in self-optimization engine. The experimental results show that despite the tremendous volume of the analyzed data – which is hundreds of times bigger than usual methods – it is possible to improve the KPIs, such as throughput, up to 30 percent by optimal resource allocation and reducing the signaling load. Also, the presence of feature extraction and parameter selection modules will reduce the response time of the self-optimization model up to 25 percent when the number of parameters is too high Moreover, numerical results indicate the superiority of using support vector machine (SVM) learning algorithm. It improves the accuracy level of decision making based on the rule-based expert system. Finally, uplink quality improvement and 15-percent increment of the coverage area under satisfied SINR conditions can be considered as outcome of the proposed scheme. Manuscript profile
• Open Access Article
  • Abstract Page
  • Full-Text
  
  4 - BSFS: A Bidirectional Search Algorithm for Flow Scheduling in Cloud Data Centers
  Hasibeh Naseri Sadoon Azizi Alireza Abdollahpouri
  10.7508/jist.2019.03.002
  To support high bisection bandwidth for communication intensive applications in the cloud computing environment, data center networks usually offer a wide variety of paths. However, optimal utilization of this facility has always been a critical challenge in a data cent More

  To support high bisection bandwidth for communication intensive applications in the cloud computing environment, data center networks usually offer a wide variety of paths. However, optimal utilization of this facility has always been a critical challenge in a data center design. Flow-based mechanisms usually suffer from collision between elephant flows; while, packet-based mechanisms encounter packet re-ordering phenomenon. Both of these challenges lead to severe performance degradation in a data center network. To address these problems, in this paper, we propose an efficient mechanism for the flow scheduling problem in cloud data center networks. The proposed mechanism, on one hand, makes decisions per flow, thus preventing the necessity for rearrangement of packets. On the other hand, thanks do SDN technology and utilizing bidirectional search algorithm, our proposed method is able to distribute elephant flows across the entire network smoothly and with a high speed. Simulation results confirm the outperformance of our proposed method with the comparison of state-of-the-art algorithms under different traffic patterns. In particular, compared to the second-best result, the proposed mechanism provides about 20% higher throughput for random traffic pattern. In addition, with regard to flow completion time, the percentage of improvement is 12% for random traffic pattern Manuscript profile