You are on page 1of 4

NATIONAL INSTITUTE OF TECHNOLOGY KARNATAKA

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

DISTRIBUTED COMPUTING SYSTEMS


Assignment

Research paper proposal on topics : 1. Min-Min Max-Min Selective Algorithm 2. Dynamic Load Balancing Algorithm 3. Fair Scheduling Algorithm

Group Members

Nishant Khandelwal (09CO62) Rahul Anand (09CO72)

Min-Min Max-Min Selective Algorithm


STUDY : The increase in supercomputer costs in one hand and the need for large-scale
computational resources on the other hand, has led to the possibility of using network of computational resources to solve large-scale problems which emerge to a method called GRID. Grid is a type of parallel and distributed system that enables the sharing, selection and aggregation of geographically distributed autonomous and heterogeneous resources dynamically at runtime depending on their availability, capability, performance, cost and users quality of service requirements. Task scheduling in grid is one of the most important technologies in grid system, which is used to schedule a task on an appropriate grid node. Task scheduling algorithms are used to improve the grid performance by minimizing the scheduling length. To make use of tremendous capabilities of the distributed system, effective and efficient scheduling algorithms are needed. Scheduling on a grid has three main phases . Phase one is resource discovery , which generates a list of potential resources. Phase two involves gathering information about those resources and choosing the best set to match the application requirements. In phase three the job is executed.

ANALYSIS : Many algorithm were proposed to obtain semi-optimal match which were
categorised into two mode. Firstly on-line mode, a task is mapped to a machine as soon as it arrives at the scheduler. MET(Minimum Execution Time) : MET assigns each task to the resource that performs it in the least amount of execution time, no matter whether this resource is available or not. MCT(Minimum Completion Time) : MCT assigns each task to the resource which obtains earliest completion time for that task. Secondly batch mode in which task are collected into a set called meta-task (MT). These sets are mapped at before scheduled times called mapping events. Min-Min: Min-Min begins with the set MT of all unassigned tasks with two phases. In the first phase, the set of minimum expected completion time for each task in MT is found. In the second phase, the task with the overall minimum expected completion time from MT is chosen and assigned to required resource. Then this task is removed from MT and the process is repeated until all tasks in the MT are mapped. Max-Min: Max-Min is very similar to Min-Min, except in phase 2. Max-Min assigns task with maximum expected completion time to the corresponding resource, in phase 2. In this algorithm expected completion time of each task is calculated and the two consecutive task time is checked with the standard deviation of overall task time. If it lies in the first half of list then long task is more in number so min-min is chosen or else it means short task is more so max-min algorithm is choosen. If lies in between then if sd is less than a certain threshold, it means the length of all tasks are in a small range, so we will select Min-Min to assign the next task or else max-min.

PROPOSAL: We would like to implement the min-min and max-min algorithm using appropriate data structures in C. REFERENCES: A Min-Min Max-Min Selective Algorihtm for Grid Task Scheduling by Kobra Etminani Dept. of Computer Engineering Ferdowsi University of Mashad Mashad, Iran and Prof. M. Naghibzadeh Dept. of Computer Engineering Ferdowsi University of Mashad Mashad, Iran ,2007

Dynamic Load Balancing Algorithm


STUDY : A phenomenon often appears in the multiprocessor system is that the load is
unbalanced. In a multi-processor system the probability of one of the processor being idle while other processor has multiple jobs queued up can be very high. Such imbalances in system load suggest that performance can be improved by either transferring jobs from the currently heavily loaded processors to the lightly loaded ones or distributing load evenly or fairly among the processors. Load on a particular node depends on the number of processes, overall configuration of the node, network bandwidth etc. To improve performance, proper load balancing algorithms are required that can make all the difference between modest performance gains and maximum performance gains. These algorithms come into two basic categories - static and dynamic. While static load balancing algorithms (SLB) take decisions regarding assignment of tasks to processors based on the average estimated values of process execution times and communication delays at compile time while Dynamic load balancing algorithms (DLB) are adaptive to changing situations and take decisions at run time.

ANALYSIS: The basic idea behind this algorithm is passing of message from one node to
the adjacent node which suggest the action to be taken. The characteristics that the multiprocessor should have are like 1.evry node has same software and hardware configuration. 2.communication mode should be duplex . 3. every node never sent a wrong information.. Primary Approach: Initially processes are stored in queue and are allotted one by one to primary nodes. Processes are migrated from heavily loaded node to light weighted node. Process migration is greatly affected by the network bandwidth and work load. Here nodes are grouped into clusters and checked within its own cluster if a light weighted node is found or else next cluster is searched and after getting a light weighted node transfer takes place as per mesage. Centralized Approach: Sometimes a light node may not be available in the current cluster and due to high congestion a far away cluster might be not reachable. So to avoid this every cluster has a centralized node which is very lightly loaded and can be loaded of any overloaded node. Thus to ensure this centralized node has better structure compare to other node present in cluster. Modified Approach: Instead of keeping a single centralised node many smaller nodes, called supporting nodes(SN) are used. These nodes are given some load initially to exploit their power or else it will be a resource waste. When anyoverloaded node requests a lightly loaded SN a comparison between the requesting process and the current process being executed at the SN is done and work is done accordingly.

Proposal: We would like to implement the modified approach algorithm using appropriate
data structures in C.

Reference: 1) Qualitative Parametric Comparison of Load Balancing Algorithms in


Distributed Computing Environment #Amit Chhabra1, Gurvinder Singh2,2006 Department of Computer Science and Engineering, Guru Nanak Dev University, Amritsar 2) An Algorithm for Dynamic Load Balancing in Distributed Systems with Multiple Supporting Nodes by Exploiting the Interrupt Service, Parveen Jain, Daya Gupta, May 2009.

Fair Scheduling Algorithm


STUDY : Grid computing has emerged as an important new field, distinguished from
conventional distributed computing by its focus on large-scale resource sharing, innovative applications, and, in some cases, high-performance orientation. An efficient use of distributed resources is highly dependent on the resource allocation by grid schedulers, where user requirements and job characteristics must be also considered. The demand for scheduling is to achieve high performance computing. The basic idea of this paper is to develop a good scheduling algorithm that can perform effectively and efficiently in terms of minimizing the error to achieve fairness and reduce the cost and time. This paper proposes a fair scheduling algorithm based on the service time error.

ANALYSIS : Main objective of this algorithm would be to improve the fairness and decrease the total completion time by minimizing the service time error. Firstly here definition of fairness are like 1.the resource should be allocated inorder of increasing demand. 2.no requesting task gets a share of resource larger than its demand.3.requesting task with unsatisfied demand get equal share of resources. Tasks with a higher demand are favored against the remaining tasks in the case of other existing algorithms which means that such tasks are given a higher priority than the others which leads to starvation that increases the completion time of tasks and no fairness is guaranteed. These issues are checked in the algorithm which alllocate resources fairly to all tasks based on the error. The process of scheduling is divided into two steps firstly,the schedulers order the task in the queue, secondly, the scheduler runs the first task in the queue for its time quantum. In this real world it is impossible to be proportional fair at all interval thus service time error is calculated where negative value indicates that a task have received less than what it deserves and a zero value indicates that it has received its ideal share. The main objective of the algorithm is to minimize the error and reduce the completion time of the tasks. A queue is created for each node and the tasks are placed in the queue for execution ,assigned a weight depending on various factors depending on the infrastructure. When the processor are free the first task in the queue is assigned to processor for particular time quantum. Later to determine as to which job will be given the next time quantum depends on the error value if the error is positive then it has been given enough of resources so it is moved to the end of the queue. The error values of the first job and the second job in the queue are compared and the job that has a lower error value gets the resource for the next time quantum. This process repeats until there are no more jobs in the queue for that node. PROPOSAL: We would like to implement the service time error algorithm using
appropriate data structures in C.

REFERENCES: A DYNAMIC ERROR BASED FAIR SCHEDULING


ALGORITHM FOR A COMPUTATIONAL GRID,2006 1 DAPHNE LOPEZ, School of Computing Sciences, VIT University, Vellore, India 632006 2 S. V. KASMIR RAJA Dean Research, SRM University, Kattankulathur, Chennai, India 603203,

You might also like