A Modified Delay Strategy for Dynamic Load Balancing in Cluster and Grid Environment

Size: px
Start display at page:

Download "A Modified Delay Strategy for Dynamic Load Balancing in Cluster and Grid Environment"

Transcription

1 A Modified Delay Strategy for Dynamic Load Balancing in Cluster and Grid Environment Hemant Kumar Mehta 1, Manohar Chandwani 2 Devi Ahilya University Indore, India 1 mehtahk@yahoo.com, 2 chandwanim1@rediffmail.com Priyesh Kanungo, Patel College of Science & Technology, Indore, India priyeshkanungo@hotmail.com Abstract This paper considers a specific cluster/ grid computing environment, where the workstations are under full load situation. In such situations, we delay the scheduling of new jobs instead of dispatching them to one of the overloaded workstations. We present a modified delay strategy mdelay, to significantly enhance delay-based scheduling algorithm proposed by Hui & Chanson and named as DELAY. The proposed mdelay algorithm produces better batch completion time and improved load balancing at workstations. The experiments are performed on a centralized scheduler. However, to improve the performance a decentralized scheduling framework based on web service or grid service is also proposed. Keywords-Cluster Computing; Delay Strategy; Dynamic Load Balancing; EcoGrid; Grid Computing I. INTRODUCTION Clusters and grids are attractive platforms for deploying parallel and distributed applications at large scale and/or for high performance. For the proper distribution of users requests, load balancing is required in the cluster and grid environment. The load balancer receives users requests and sends them to specific workstations in the cluster or grid according to some distribution logic. Generally, load balancing enables effective allotment of bottleneck resources to improve the overall performance of the system. Load balancing (LB) can mainly be classified as static, dynamic and adaptive. In case of static load balancing, the decision regarding the allotment of a request to a resource is known well in advance. In case of dynamic load balancing (DLB), the request allotment decisions are taken at run time based on the current state of the system. In adaptive load balancing (ALB), the parameters and policies can change depending on the run time information about the previous decisions, their results and the current state of the system [2][ 8]. According to Hui and Chanson [5], most of the dynamic load balancing algorithms ignore several practical issues. Generally, the scheduling algorithms immediately dispatch the jobs upon their arrival even if the system is under full utilization. If a batch with large number of jobs is arrive in the full load condition, then both the existing and newly arrived processes suffer from poor response time. Accordingly, the execution of newly arrived job must be delayed until one of the currently running processes terminates. This improves overall performance of the system. The DELAY strategy is generic and can be incorporated in any existing scheduling algorithm. The DELAY strategy enables the system to adopt sudden arrival of large batch and to provide the equal treatment to all the processes thereby improving the job fairness. Generally, application having high computation requirement are deployed over cluster and grid environments. The cluster and grid environment used for such applications are under overloaded situation. Therefore the DELAY strategy is suitable to such an environment. The scheduler used for the experiments is a centralized scheduler. The centralized scheduler is simple to implement. However, the centralized scheduler may results in single point failure and the limited scalability in terms of performance as there are a large number of resources in the grid. The collection of load status on these workstations must be a frequent operation causing the degraded performance. The problem may be triumph over using a decentralized load status collection model. This paper also proposes a web service or grid service based framework for the decentralized load status collection. As per our literature survey, the DELAY strategy proposed by Hui and Chanson [5] is not updated by any other researcher except the work by Zhang and Lu. However, they have directly applied the strategy to grid without any modification to the delay strategy proposed by Hui and Chanson. However, to collect load information from large number of resources in the grid, they proposed a mobile agent based framework for collecting the load information from the various nodes and submitting to centralize scheduler [9]. This paper is organized as follows. Section 2 examines related work and the proposed modifications. Section 3 describes the mdelay algorithm. The simulator used for experimentation and the experimental setup is described in Section 4. In Section 5, we presented and analyzed the results of experiments. Section 6 presents a web service or grid service based two level framework for the load collection. Finally, Section 7 concludes our research and development effort. II. THE DELAY STRATEGY AND THE NEW MODIFICATIONS DELAY strategy uses scheduling function to reflect the desirability of executing a new job on a particular workstation. A new job is not directly submitted to an already saturated system. In such a scenario, the new job waits for completion of some of the running jobs. The scheduling function is derived on the basis of the bounding factor of the workstations, the total number of jobs currently running under the workstation /10/$ IEEE

2 and the speed of the processor. The bounding factors are predetermined and statically allotted to the workstations. The DELAY algorithm takes the bounding factors and speed of the processors as input. The experiments are performed in a controlled environment comprising of seven Sun SPARC workstations connected in LAN environment. The relative speed is assigned to each of the workstations [5]. Hui and Chanson have not mentioned the basis of assigning the relative speed. We have computed the relative speed on the basis of Dhrystone benchmark. The benchmark provides MIPS to each processor on the basis of the speed of the processor. In our research, we have considered processors with the values that are multiple of Dhrystone benchmark MIPS of their previous processors as shown in Table I [7]. Most of the scheduling algorithms use the run queue length in order to describe the workload on the system. The run queue length represents the total number of jobs currently running or they are in ready queue. But according to Hui and Chanson, this parameter is not enough, as in a LAN environment, there may be several jobs using co-processors of the workstations. These jobs will be in the blocked state, hence they will not be considered in the list of running processes. However, they may be currently executing on any of the co-processors of the workstation. Therefore, Hui and Chanson have taken another metric called number of active jobs on the workstation. The number of active jobs includes the run queue length as well as the jobs running or waiting for co-processors of the workstations. As we are mainly concerned with the main processor of the workstations instead of the metric taken by Hui and Chanson, we choose another metric called number of effective processes [3]. Number of effective processes means the number of processes either currently running or they are ready to use the main processor of the workstation but not the coprocessor of the system. The value of parameter is computed on the basis of workload characteristics [7] i.e. CPU-bound workload and I/O-bound workload. If there is an overlap between the I/O-bound workload and CPU-bound workload, there is a possibility that the total number of effective jobs is much less than the total number of jobs in the system. Therefore, while calculating the load on the system, we are counting the load on the main processor only and excluding the load on the coprocessors. Moreover, DELAY algorithm doesn t consider the execution time of the jobs while determining the workstation load as this value is difficult to estimate. In this case, there is a strong possibility that several long jobs may be allocated to the same workstation, resulting in poor batch completion time as depicted in the Fig. 1. According to Krishnaswamy, Loke & Zaslavsky and Ali et al., the runtime of a job can be predicted with 80% accuracy in several situations [1] [6]. There are several situations where execution time of a job can be estimated. These situations can be conversion of flat file into database, file compression/ decompression, exporting the data from database etc. Depending on the number of records to be processed and the processor type, the execution time of such jobs can be estimated. Accurate estimation of execution time can improve the performance of DELAY strategy. We have considered the execution time of job as a parameter along with number of effective processes, to determine the current workload. The value of bounding factor is predetermined in case of DELAY strategy. But if this value is not determined properly, the utilization of the processor will be poor. Also in the adaptive delay algorithm (DELAY*) by Hui and Chanson, the value of bounding factor is just initialized to 1, the value is updated based on the length of pending queue. However, the pending queue length can not be the optimal measure of the processor load as it also dependent on the execution time of the jobs. In mdelay we consider the execution time of job along with the queue length to compute the bounding factors of the workstations. Hui and Chanson have taken several performance matrices in order to show the effectiveness of the DELAY strategy. Various matrices taken by them are: Batch Completion time (B): The batch completion time is defined as the time elapsed in the execution of the entire batch of the work. This metric represents the overall throughput of the system. Proper division of load among the nodes results in the smaller value of B. The smaller value represents the higher throughput and even distribution of the workload. Overall speedup (S): The S is the ratio of B with the batch completion time of the same batch when it is executed on the slowest workstation. Mean response time ( Τ): The mean response time is the average value of completion time of processes in a batch. Mean job completion time ( J ): The job completion time (J) is defined as the ratio of job execution in normal situation with the execution time of the job when it is executed on the fastest idle machine. The average value of J is mean job completion time. Standard deviation of J (sd( J )): The standard deviation of job completion time denotes the fairness of the scheduling algorithm. Average waiting period ( W): The waiting period W is the percentage of the waiting time of the job with the execution time. Parameters B & S indicate the throughput of the system, Τ & J indicate the performance of individual jobs and sd(j) & W represent the fairness of the scheduling algorithms. TABLE I. THE CONFIGURATION,DHRYSTONE BENCHMARK AND RELATIVE SPEED OF THE PROCESSORS Processor Speed in MHz Dhrystone VAX MIPS Relative Speed Pentium III Pentium P4 Xeon Pentium 4E Pentium Core 2 Duo 1 CP

3 Q u e u e Scheduler Workstations :- Longer Job :- Normal Job Figure 1. Several longer jobs allocated to the same node Instead of taking all these parameters for the comparison, we have considered only two parameters to represent the fairness and throughput of the scheduling algorithms. The parameters are batch completion time and workstation processing completion time. If workstations are completing their processing in less time, then throughput of the algorithm is higher. Similarly, if batch completion time is smaller then each job is getting proportionate CPU time resulting in the improved fairness of the algorithm. We have also plotted the graphs for comparing the variation among these values. From the graphs, it is observed that the matrices considered by us can very well replace the performance matrices considered by Hui and Chanson. III. DESCRIPTION OF THE ALGORITHM: MDELAY The proposed mdelay strategy is implemented in Java. The scheduling process starts with the creation of cluster and grid followed by assignment of relative speed and bounding factor to the workstations. Initially, as soon as the batches arrive, the algorithm determines the best suitable workstation for each of the processes and dispatches processes one at a time to the workstation immediately. This immediate dispatching process continues till the load of all workstations is below the full load capacity. When the load of all the workstation reaches to the full load capacity, then the algorithm starts delaying the processes till one of the currently executing process terminates. While determining the load, the execution time of job is also taken into account along with the run queue length. The execution capability of each workstation is determined by its speed in terms of Dhrystone benchmark MIPS. In order to test and compare the proposed modifications, we have implemented and tested the following three algorithms: Distributed Round Robin: This is most popular scheduling algorithm used in processor scheduling. This algorithm executes processes for a fixed time slice then switches the context to another process. This process continues till the completion of all the processes [4]. DELAY strategy: This is the strategy proposed by Hui and Chanson. mdelay strategy: The DELAY strategy modified by authors of this paper. Distributed round robin algorithm is one of the most generic and simple algorithms. Distributed round robin, DELAY and mdelay algorithms have been implemented in Java. The implementation uses the facility provided by the simulator to create the cluster/ grid. The algorithm for mdelay strategy is mentioned below: Algorithm mdelay: 1. Create and initialize the cluster or grid on EcoGrid. 2. Assign the relative speed to the processors of workstation 3. Assign bounding factor for each of the workstations based on the relative speed calculated in Step Step 4 and Step 5 will be repeated until processes in all the batches terminates a. Fetch a process from the batch of processes arrived at the cluster/ grid b. Allocate suitable workstation to the processes c. If the workload of the workstation is less than its capability, then immediately dispatch the process to the workstation d. Else delay the process till one of the currently executing process terminates 5. for each workstation in the cluster/ grid, repeat the Step 5 (a) to Step 5 (c) a. Fetch a process from workstation queue b. Execute the process for the allocated time quantum c. Update the execution parameter for the process 6. Go to step 4 End of Algorithm IV. EXPERIMENTATION All algorithms are implemented and tested on a Java based simulator called EcoGrid. EcoGrid provides support for the creation of the cluster and grid as well as execution of the processes on it. EcoGrid provides an interface GridScheduler which is inherited in the classes which are defined for the scheduling algorithms. This grid simulator is developed as a test-bed for grid scheduling algorithms that are based on dynamic load balancing. EcoGrid supports simulation of the economy-based as well as non-economy based scheduling algorithm. A. Architecture of EcoGrid The grid environment supported in EcoGrid contains various components required in the grid. The architecture of EcoGrid is shown in Fig. 2. Different components of EcoGrid are: Configuration Manager (CM), Random Number Generator (RNG), Load Generator (LG), Resource Calendar (RC),

4 Computer Node (CN), Computer Cluster (CC), Media Directory (MD), Grid Process (GP), Grid (G), Grid Scheduler (GS), Statistical Analyzer (SA) and the Grid Data Provider (GDP). The DB & FS represents database and file system. The components of EcoGrid are being described below: 1) Configuration Manager (CM): The simulator is dynamically configurable. Configuration manager is responsible for dynamically configuring the grid simulator. Several methods have been included to read and update the values of the parameters. 2) Random Number Generator (RNG): This component provides the methods for random number generation that are based on the various statistical distributions namely Poisson distribution, Exponential distribution, Normal distribution and Uniform distribution. 3) Resource Calendar (RC): The resource calendar element of the simulator is used by the computer node to define its basic properties. It stores the details regarding the load, price, booking status, list of holidays of the time zone etc. for the resource. The load and price are having different values for weekdays and for holidays. 4) Load Generator (LG): This module generates the processing load for the system. We can also configure the parameters of the load generator for varying work load situations. 5) Statistical Analyzer (SA): The statistics generated by the system is analyzed by this component. This module provides the facility to perform statistical computations. 6) Grid Data Provider (GDP): This component is designed to separate the data access functionality from the core grid functionality. The class will be used to facilitate data access from various sources. Data can be stored in flat files, database software or XML files. 7) Computer Node (CN): The basic resource in the grid is represented by the computer node component. This component consists of the hard disk, main memory and the processor. Based on the speed of the processor, a weight is assigned to the computer node, so that it can represent higher class of the processing unit as compare to the computer nodes with normal weight (=1). Figure 2. Architecture of EcoGrid 8) Computer Cluster (CC): The computer cluster is the collection of the computer nodes to represent the processing unit of the grid. Cluster will publish its details in the media directory to facilitate the resource consumers. 9) Media Directory (MD): As the name suggests, this component acts as the media service to store the details of the service providing clusters, which can be accessed by the resource consumers. It also provides resource matching algorithm to determine most suitable cluster and most suitable node. Media directory contains two databases: mediafileprovider containing the details of the service providing cluster and the mediafileprocess containing the details of the negotiated processes. 10) Grid (G): The Grid represents a set of cluster to perform the task. It initializes the clusters, nodes in the cluster and the various resources on a node. 11) Grid Process (GP): Grid process represents a task to be executed on grid. The arrival time for the process follows the Poisson distribution. The details required regarding the process are: execution time (executiontime), input data file size (inputfilesize), starting date (arrivaldate), deadline date (deadlinedate), deadline miss limit (deadlinemisslimit) budget, optimization factor (optimizationfactor) and negotiation factor for time & cost (timenegotiationfactor & costnegotiationfactor). 12) Grid Scheduler (GS): Grid scheduler represents the set of algorithms to schedule the processes on the clusters and the nodes. We have designed the grid scheduler in such a manner that the users can dynamically plug their own scheduling algorithms into the system. B. Configurable Parameters Supported in EcoGrid In EcoGrid, we can configure several parameters dynamically. The essential details about the components of the grid which may require changes are taken as configurable parameter. For example, the list of holidays, the load and price detail of a computer node, the number of clusters in the grid, name of the databases of the media directory etc. Various configurable parameters are described in Table II which may be defined as per the needs of the users. TABLE II. VARIOUS CONFIGURABLE PARAMETERS OF ECOGRID Parameter Description The following parameters can be defined based on the experience of the system: weekdayload Defines load of the 24 hours of a weekday for the node. holidayload Defines the load of 24 hours of a holiday for the node. holidays Contains list of holidays for the particular zone. weekdayprice The price of the node. It contains the different price values for the different hours of the weekday. holidayprice The price of the node. It contains the different price values for the different hours of the holiday. specialoffers Contains details about the special offers, including the date, hour and price for the hour.

5 reservedays The number of days for the advance reservation. mediafileprovider The name of media database which contains the billboard of clusters. mediafileprocess The name of the media database which contains the details of processes. lambda The mean parameter for the Poisson distribution. deadlinemisslimit The extra time given to the process to finish after the deadline date. costnegotiationfactor The factor by which the cost of a process can be negotiated. timenegotiationfactor This factor represents the number of days by which the deadline of the process can be extended. optimizationfactor Decides whether the process is optimized by cost, time or both. schedulingalgorithm The name of the class having the scheduling algorithm to use. schedulartype Indicates that whether the scheduling algorithm is economy based on not. The following parameters are generated as random numbers between the given range: hddspace Represents the hard disk size of a node. memory Main memory of a node. cpuspeed The speed of the CPU of the node. nodespergridunit The number of nodes in a grid unit (cluster). executiontime The execution time of the process. inputfilesize The input file size required by the grid process. arrivaldate The value represents the arrival date of the process. The arrival date follows the Poisson distribution. deadlinedate The date before which the process must be completed. Budget The budget of the process. cpubound Percentage of CPU bound processes. clusterpergrid Number of clusters in grid. All the components of EcoGrid use each other. The creation of grid will begin from the creation of CN and CC thereafter CC and CN initializes them self. CN also initializes RC. After initialization of the grid, the LG starts generating the GPs to be processed at the grid. The GS and MD schedule these GPs to various CN available in the grid. The SA, GDP, RNG and CM are utility modules used by the other components grid. C. Experimental Setup Two experiments are performed using the EcoGrid simulation environment. In the first experiment six workstations have been considered. As shown in the Table I, The workstations have their Dhrystone MIPS values as the multiple of the MIPS value of previous workstations. In order to test the effect of execution time of jobs, we have created one large process in each of the batch arriving at system. The batch size is dynamically configurable and kept fixed for a given experiment. In the experiment, the size of normal batches is seven. The size of delayed batches is ten. The second experiment is performed to test algorithm on an environment consisting of large number of workstations. In this experiment total hundred workstations are considered. The size of normal batches is eighty and the size of delayed batches is three hundred. Six different types of processors are considered as shown in the Table I. In order to test the effect of execution time of jobs, we have created one large process in each of the normal batches and seven large processes in each of the delayed batches arriving at system. Generally, the batches of the processes are created until the workload reaches to the full capacity of the system. After the full load situation, few special batches with a large number of jobs are created. These batches are kept in another pool, and a job is extracted from this pool when one of the executing processes terminates. The generated workload is having two kinds of jobs namely CPU bound jobs and I/O bound jobs. The simulator used for testing the algorithms is a dynamically configurable. We defined a parameter to determine the ratio of the CPU bound processes and I/O bound processes. For the current execution, the ratio of number of CPU bound jobs to I/O bound job is 80:20. The results of the experiments have been analyzed in the next section. V. INTERPRETATION OF RESULTS The results of the first experiments are shown in Fig. 3 and Fig. 4 respectively. We observe variation in the completion time of the batches of the jobs and the last process at workstations. The variation represents the balancing state of the system. The smaller value of variation denotes the better load balancing at the workstation and better response time for the batches. Fig. 3 represents that when a particular workstation has finished the last candidate process. By the termination time of workstation we mean the time by which all the processes allocated to that workstation get completed. The most balanced state is a state when all the workstations complete their processing with minimal variation. From the graph in Fig. 3, we can observe that most balanced state among these three algorithms can be achieved by our proposed mdelay strategy. We can also observe that the average completion time of the last process at each workstation is lowest for mdelay strategy. In the present example, the processing at workstations using mdelay algorithm, completes 20% earlier than DELAY strategy. Termination Time Stamp of a Workstaion Workstation-id Round Robin Average:361 DELAY Average:264 mdelay Average:213 Figure 3. Variation in Completion Time of Processes at Various Workstations using Distributed Round Robin, DELAY and mdelay

6 Termination Time Stamp of a Batch Round Robin Average:355 DELAY Average:213 mdelay Average:177 Terrmination Time Stamp of a Workstation Round Robin DELAY mdelay Batch # Experiments with the Hundred Workstations Figure 4. Variation in the Completion Time of Various Batches using Distributed Round Robin, DELAY and mdelay. Fig. 4 shows the scheduling status of various batches. By the termination time of batch we mean the time when all the processes of a batch complete their execution. We observe that the average batch completion time and variation in batch completion time is again the smallest using proposed mdelay algorithms. The present example also shows that the average batch completion time of mdelay strategy is around 17% less than DELAY strategy. From the results given above, it can also be interpreted that the mdelay algorithm improves the job fairness, as the variation in the batch completion time is minimum in case of mdelay strategy. The smaller value of variation in the termination time of first and last process denotes that they have been given fair treatment. Another observation from Fig. 3 and Fig. 4 is that, using mdelay strategy, activities are finished with minimum variation in their minimum and maximum values at a workstation. This variation represents balanced state among the workstations because they are completing their processing with the minimum variation. In the second experiment the data is collected from the two hundred iterations of the experiment. In one repetition average of completion time of last candidate process at each of the workstations is calculated. At the same time, the average of the completion time of the last candidate process from each of the batches is also calculated. Fig. 5 and Fig. 6 are drawn from the results of the second experiment and confirm the result from the first experiment. The mdelay produces 16% improvement in average completion time of last process at each workstation and 10 % improvement in average completion time of each batch. The percentage of improvement is less in second experiment as it is having an average value of two hundred iterations while the first experiment produces the result of single execution. While performing the experiments we also considered different combinations of number of nodes, job sizes and the batch sizes for normal batches and long batches. However, we observe that the results do not depend on any of these factors. Figure 5. Average Completion times of Processes of all Workstations using Distributed Round Robin, DELAY and mdelay Termination Time Stamp of a Batch Figure 6. Average Completion Time of Various Batches using Distributed Round Robin, DELAY and mdelay VI Experiments with the Hundred Workstations Round Robin Average: 6825 DELAY Average: 1784 mdelay Average: 1628 A SERVICE BASED DISRIBUTED LOAD COLLETION FRAMWORK A. Web Service The web services are the software designed to support the interaction between different machines connected through the network in a loosely coupled client server manner. The service requesters can only be software that requires information provided by the web service. Simple Object Access Protocol (SOAP) is used for preparation of the request to and response from the web services. The request and response messages are prepared as per the specification of SOAP web services publish their details to a registry called Universal Description, Discovery and Integration (UDDI). UDDI is a platform independent registry for web services. UDDI uses the Web Service Definition Language (WSDL) to describe the details of web services. Using WSDL the detailed functionality of web services can be specified. These details are accessed by the clients to invoke the functionality of web services. Web services use the existing internet infrastructure and HTTP for the interaction with its requesters. Extensible Markup Language (XML) is a standard to specify the data about the

7 applications. SOAP messages, UDDI and WSDL use XML for specifying their details. The web services provide good solution for a cluster but due to its limitation it is not much suitable for grid applications. The main limitation of a web service is that the web services are stateless and persistent services. The web service is stateless means it doesn t store data of the same client between different requests. Consider a task that cannot be accomplished in a single operation but it requires several operations. To complete such task the web service must store the data of the clients between its subsequent requests. Another related problem of web service is its persistence, if a web service persist the data of a different requests of a client than the same data is visible to other clients. However, the data must be transient between the users. The grid applications demands the web services to be transient and yet stateful. To overcome these problems the concept of grid services is introduced. B. Grid Service In last few years, users have started harnessing the power of web services for grid computing called service oriented grid computing or grid services. The grid services are stateful and transient in nature. The largest grid user s community called Open Grid Forum (OGF) has described architecture that is standard and common to all grid services for commercial and scientific applications. This architecture is known as Open Grid Services Architecture (OGSA). OGSA takes care of heterogeneity of the resources resulting in the proper communication and sharing of the information between them. However, OGSA is a general specification that outlines what a grid service should have in a broad manner. The Open Grid Services Infrastructure (OGSI) a part of OGSA standard that provides technical specification of the grid service. VII. TWO-LEVEL SERVICE BASED FRAMEWORK The presented implementation of mdelay strategy is based on centralized scheduler. The centralized scheduler is considered in order to compare the performance of mdelay and DELAY strategy. However, the centralized scheduler lacks up-to-date load status of each workstation in the grid or suffers from the degraded performance due to the high frequency of load status information. To cope up with the problem a service based decentralize framework is proposed. Moreover, the mobile agent based decentralized framework proposed by Zhang and Lu can be effectively used in cluster computing environment. However, the mobile agent based framework is not suitable for the grid computing environment. Grid contains a large number of resources and the single mobile agent can not effective collect the load information from such a large set of resources in the grid. Effectively this model is more or less same as the centralized scheduling model. Hence we proposed the new model to improve the process of load collection. As depicted in the Fig. 7, the two-level grid services are proposed to form a decentralized environment. The framework considers hierarchical grid environment where each workstation belongs to a cluster and grid contains several such clusters. The level one service is the single service defined at grid level. It is the called master service that provides up-todate workstation load information to the scheduler. The master service uses level two services to have the recent workstation load information. The level-two service is called as slave service. One slave service is attached with each cluster that belongs to the grid. The local service of each cluster collects the load information of each workstation of the cluster and pushes the information to the master service. In this manner the latest load status of each workstation will be available to the scheduler. This decentralized workstation load status collection framework is more effective than the framework proposed by Zhang and Lu as in mobile agent based framework the information is sent individually for each workstation whereas in the proposed framework cluster wide load status is sent. Cluster 1 Collection Master Service Collection Cluster 3 Slave Service 1 Slave Service 3 Submission Submission Collection Slave Service 2 Scheduler Slave Service n Collection Cluster 2 Cluster n Figure 7. Two-Level Service Based Decentralize Framework for mdelay Strategy

8 VIII. CONCLUSION In this paper, we proposed modifications in the DELAY strategy by Hui and Chanson in several ways to obtain shorter execution schedule resulting in improved load balancing at workstations in a cluster/ grid. The new modified delay strategy is referred as mdelay. Using mdelay strategy the workstations finish their processing with minimal variation from the average values. The batch completion time is also having minimal variation from the average values. A reduction of average 10% in batch completion time and average 16% reduction in workstation processing completion time have been observed. A two-level service based decentralized framework is also proposed to implement the mdelay scheduling strategy for improved performance over the centralized scheduler. REFERENCES [1] A. Ali et al., Pedicting the resource requirements of a job submission, International Conference on Computing in High Energy and Nuclear Physics, [2] T.L. Casavant and J.G. Kuhl, A taxonomy of scheduling in general purpose distributed computing system, IEEE Transaction on Software Engineering, Vol. 14, No. 2, February [3] M. Choi, J. Yu and S. Maeng, Improving performance of a dynamic load balancing system by using number of effective tasks, KISS Computer System Conference, October [4] D. McLaughlin, S. Sardesai and P. Dasgupta, Preemptive scheduling for distributed systems, 11th International Conference on Parallel and Distributed Computing Systems, September [5] C.C. Hui and S.T. Chanson, Improved strategies for dynamic load balancing, IEEE Concurrency, Vol. 7, July [6] S. Krishnaswamy, S.W. Loke, A. Zaslavsky Estimating computation times of data-intensive applications, IEEE Distributed Systems Online, Vol. 5, No. 4, April [7] Results of dhrystone benchmark for CPUs. #anchorcpu (accessed on 7 July 2009). [8] H. Mehta, P. Kanungo and M. Chandwani, Performance enhancement of scheduling algorithms in web server clusters using improved dynamic load balancing policies, 2nd National Conference, INDIACom-2008 Computing For Nation Development, New Delhi, February 2008, pp [9] J. Zhang and X. Lu, A Dynamic Job Scheduling Algorithm for Computational Grid, Second International Workshop Grid and cooperative computing:, December 2003, pp

Performance Enhancement of Scheduling Algorithms in Clusters and Grids using Improved Dynamic Load Balancing Techniques

Performance Enhancement of Scheduling Algorithms in Clusters and Grids using Improved Dynamic Load Balancing Techniques Performance Enhancement of Scheduling Algorithms in Clusters and Grids using Improved Dynamic Load Balancing Techniques Hemant Kumar Mehta School of Computer Science, Devi Ahilya University, Indore, India

More information

CS370: System Architecture & Software [Fall 2014] Dept. Of Computer Science, Colorado State University

CS370: System Architecture & Software [Fall 2014] Dept. Of Computer Science, Colorado State University Frequently asked questions from the previous class survey CS 370: SYSTEM ARCHITECTURE & SOFTWARE [CPU SCHEDULING] Shrideep Pallickara Computer Science Colorado State University OpenMP compiler directives

More information

Journal of Global Research in Computer Science

Journal of Global Research in Computer Science Volume 2, No. 4, April 211 Journal of Global Research in Computer Science RESEARCH PAPER Available Online at www.jgrcs.info A New Proposed Two Processor Based CPU Scheduling Algorithm with Varying Time

More information

A COMPARATIVE STUDY OF CPU SCHEDULING POLICIES IN OPERATING SYSTEMS

A COMPARATIVE STUDY OF CPU SCHEDULING POLICIES IN OPERATING SYSTEMS VSRD International Journal of Computer Science &Information Technology, Vol. IV Issue VII July 2014 / 119 e-issn : 2231-2471, p-issn : 2319-2224 VSRD International Journals : www.vsrdjournals.com REVIEW

More information

Chapter 3. Design of Grid Scheduler. 3.1 Introduction

Chapter 3. Design of Grid Scheduler. 3.1 Introduction Chapter 3 Design of Grid Scheduler The scheduler component of the grid is responsible to prepare the job ques for grid resources. The research in design of grid schedulers has given various topologies

More information

CPU Scheduling. CSE 2431: Introduction to Operating Systems Reading: Chapter 6, [OSC] (except Sections )

CPU Scheduling. CSE 2431: Introduction to Operating Systems Reading: Chapter 6, [OSC] (except Sections ) CPU Scheduling CSE 2431: Introduction to Operating Systems Reading: Chapter 6, [OSC] (except Sections 6.7.2 6.8) 1 Contents Why Scheduling? Basic Concepts of Scheduling Scheduling Criteria A Basic Scheduling

More information

Dynamic Load balancing for I/O- and Memory- Intensive workload in Clusters using a Feedback Control Mechanism

Dynamic Load balancing for I/O- and Memory- Intensive workload in Clusters using a Feedback Control Mechanism Dynamic Load balancing for I/O- and Memory- Intensive workload in Clusters using a Feedback Control Mechanism Xiao Qin, Hong Jiang, Yifeng Zhu, David R. Swanson Department of Computer Science and Engineering

More information

ADAPTIVE AND DYNAMIC LOAD BALANCING METHODOLOGIES FOR DISTRIBUTED ENVIRONMENT

ADAPTIVE AND DYNAMIC LOAD BALANCING METHODOLOGIES FOR DISTRIBUTED ENVIRONMENT ADAPTIVE AND DYNAMIC LOAD BALANCING METHODOLOGIES FOR DISTRIBUTED ENVIRONMENT PhD Summary DOCTORATE OF PHILOSOPHY IN COMPUTER SCIENCE & ENGINEERING By Sandip Kumar Goyal (09-PhD-052) Under the Supervision

More information

LECTURE 3:CPU SCHEDULING

LECTURE 3:CPU SCHEDULING LECTURE 3:CPU SCHEDULING 1 Outline Basic Concepts Scheduling Criteria Scheduling Algorithms Multiple-Processor Scheduling Real-Time CPU Scheduling Operating Systems Examples Algorithm Evaluation 2 Objectives

More information

Improved Task Scheduling Algorithm in Cloud Environment

Improved Task Scheduling Algorithm in Cloud Environment Improved Task Scheduling Algorithm in Cloud Environment Sumit Arora M.Tech Student Lovely Professional University Phagwara, India Sami Anand Assistant Professor Lovely Professional University Phagwara,

More information

A Study on Load Balancing Techniques for Task Allocation in Big Data Processing* Jin Xiaohong1,a, Li Hui1, b, Liu Yanjun1, c, Fan Yanfang1, d

A Study on Load Balancing Techniques for Task Allocation in Big Data Processing* Jin Xiaohong1,a, Li Hui1, b, Liu Yanjun1, c, Fan Yanfang1, d International Forum on Mechanical, Control and Automation IFMCA 2016 A Study on Load Balancing Techniques for Task Allocation in Big Data Processing* Jin Xiaohong1,a, Li Hui1, b, Liu Yanjun1, c, Fan Yanfang1,

More information

Today s class. Scheduling. Informationsteknologi. Tuesday, October 9, 2007 Computer Systems/Operating Systems - Class 14 1

Today s class. Scheduling. Informationsteknologi. Tuesday, October 9, 2007 Computer Systems/Operating Systems - Class 14 1 Today s class Scheduling Tuesday, October 9, 2007 Computer Systems/Operating Systems - Class 14 1 Aim of Scheduling Assign processes to be executed by the processor(s) Need to meet system objectives regarding:

More information

Enhanced Round Robin Technique with Variant Time Quantum for Task Scheduling In Grid Computing

Enhanced Round Robin Technique with Variant Time Quantum for Task Scheduling In Grid Computing International Journal of Emerging Trends in Science and Technology IC Value: 76.89 (Index Copernicus) Impact Factor: 4.219 DOI: https://dx.doi.org/10.18535/ijetst/v4i9.23 Enhanced Round Robin Technique

More information

CPU Scheduling. Operating Systems (Fall/Winter 2018) Yajin Zhou ( Zhejiang University

CPU Scheduling. Operating Systems (Fall/Winter 2018) Yajin Zhou (  Zhejiang University Operating Systems (Fall/Winter 2018) CPU Scheduling Yajin Zhou (http://yajin.org) Zhejiang University Acknowledgement: some pages are based on the slides from Zhi Wang(fsu). Review Motivation to use threads

More information

Uniprocessor Scheduling. Basic Concepts Scheduling Criteria Scheduling Algorithms. Three level scheduling

Uniprocessor Scheduling. Basic Concepts Scheduling Criteria Scheduling Algorithms. Three level scheduling Uniprocessor Scheduling Basic Concepts Scheduling Criteria Scheduling Algorithms Three level scheduling 2 1 Types of Scheduling 3 Long- and Medium-Term Schedulers Long-term scheduler Determines which programs

More information

A Resource Discovery Algorithm in Mobile Grid Computing Based on IP-Paging Scheme

A Resource Discovery Algorithm in Mobile Grid Computing Based on IP-Paging Scheme A Resource Discovery Algorithm in Mobile Grid Computing Based on IP-Paging Scheme Yue Zhang 1 and Yunxia Pei 2 1 Department of Math and Computer Science Center of Network, Henan Police College, Zhengzhou,

More information

Scheduling of processes

Scheduling of processes Scheduling of processes Processor scheduling Schedule processes on the processor to meet system objectives System objectives: Assigned processes to be executed by the processor Response time Throughput

More information

Network Load Balancing Methods: Experimental Comparisons and Improvement

Network Load Balancing Methods: Experimental Comparisons and Improvement Network Load Balancing Methods: Experimental Comparisons and Improvement Abstract Load balancing algorithms play critical roles in systems where the workload has to be distributed across multiple resources,

More information

A QoS Load Balancing Scheduling Algorithm in Cloud Environment

A QoS Load Balancing Scheduling Algorithm in Cloud Environment A QoS Load Balancing Scheduling Algorithm in Cloud Environment Sana J. Shaikh *1, Prof. S.B.Rathod #2 * Master in Computer Engineering, Computer Department, SAE, Pune University, Pune, India # Master in

More information

Grid Computing Fall 2005 Lecture 5: Grid Architecture and Globus. Gabrielle Allen

Grid Computing Fall 2005 Lecture 5: Grid Architecture and Globus. Gabrielle Allen Grid Computing 7700 Fall 2005 Lecture 5: Grid Architecture and Globus Gabrielle Allen allen@bit.csc.lsu.edu http://www.cct.lsu.edu/~gallen Concrete Example I have a source file Main.F on machine A, an

More information

Performance Extrapolation for Load Testing Results of Mixture of Applications

Performance Extrapolation for Load Testing Results of Mixture of Applications Performance Extrapolation for Load Testing Results of Mixture of Applications Subhasri Duttagupta, Manoj Nambiar Tata Innovation Labs, Performance Engineering Research Center Tata Consulting Services Mumbai,

More information

Preview. Process Scheduler. Process Scheduling Algorithms for Batch System. Process Scheduling Algorithms for Interactive System

Preview. Process Scheduler. Process Scheduling Algorithms for Batch System. Process Scheduling Algorithms for Interactive System Preview Process Scheduler Short Term Scheduler Long Term Scheduler Process Scheduling Algorithms for Batch System First Come First Serve Shortest Job First Shortest Remaining Job First Process Scheduling

More information

Uniprocessor Scheduling. Aim of Scheduling

Uniprocessor Scheduling. Aim of Scheduling Uniprocessor Scheduling Chapter 9 Aim of Scheduling Response time Throughput Processor efficiency Types of Scheduling Long-Term Scheduling Determines which programs are admitted to the system for processing

More information

Uniprocessor Scheduling. Aim of Scheduling. Types of Scheduling. Long-Term Scheduling. Chapter 9. Response time Throughput Processor efficiency

Uniprocessor Scheduling. Aim of Scheduling. Types of Scheduling. Long-Term Scheduling. Chapter 9. Response time Throughput Processor efficiency Uniprocessor Scheduling Chapter 9 Aim of Scheduling Response time Throughput Processor efficiency Types of Scheduling Long-Term Scheduling Determines which programs are admitted to the system for processing

More information

International Journal of Advanced Research in Computer Science and Software Engineering

International Journal of Advanced Research in Computer Science and Software Engineering Volume 3, Issue 4, April 2013 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Reducing the Number

More information

A SURVEY ON SCHEDULING IN HADOOP FOR BIGDATA PROCESSING

A SURVEY ON SCHEDULING IN HADOOP FOR BIGDATA PROCESSING Journal homepage: www.mjret.in ISSN:2348-6953 A SURVEY ON SCHEDULING IN HADOOP FOR BIGDATA PROCESSING Bhavsar Nikhil, Bhavsar Riddhikesh,Patil Balu,Tad Mukesh Department of Computer Engineering JSPM s

More information

Process- Concept &Process Scheduling OPERATING SYSTEMS

Process- Concept &Process Scheduling OPERATING SYSTEMS OPERATING SYSTEMS Prescribed Text Book Operating System Principles, Seventh Edition By Abraham Silberschatz, Peter Baer Galvin and Greg Gagne PROCESS MANAGEMENT Current day computer systems allow multiple

More information

Load Balancing in Cloud Computing System

Load Balancing in Cloud Computing System Rashmi Sharma and Abhishek Kumar Department of CSE, ABES Engineering College, Ghaziabad, Uttar Pradesh, India E-mail: abhishek221196@gmail.com (Received on 10 August 2012 and accepted on 15 October 2012)

More information

An Improved Priority Dynamic Quantum Time Round-Robin Scheduling Algorithm

An Improved Priority Dynamic Quantum Time Round-Robin Scheduling Algorithm An Improved Priority Dynamic Quantum Time Round-Robin Scheduling Algorithm Nirali A. Patel PG Student, Information Technology, L.D. College Of Engineering,Ahmedabad,India ABSTRACT In real-time embedded

More information

OVERHEADS ENHANCEMENT IN MUTIPLE PROCESSING SYSTEMS BY ANURAG REDDY GANKAT KARTHIK REDDY AKKATI

OVERHEADS ENHANCEMENT IN MUTIPLE PROCESSING SYSTEMS BY ANURAG REDDY GANKAT KARTHIK REDDY AKKATI CMPE 655- MULTIPLE PROCESSOR SYSTEMS OVERHEADS ENHANCEMENT IN MUTIPLE PROCESSING SYSTEMS BY ANURAG REDDY GANKAT KARTHIK REDDY AKKATI What is MULTI PROCESSING?? Multiprocessing is the coordinated processing

More information

Chapter 6: CPU Scheduling. Operating System Concepts 9 th Edition

Chapter 6: CPU Scheduling. Operating System Concepts 9 th Edition Chapter 6: CPU Scheduling Silberschatz, Galvin and Gagne 2013 Chapter 6: CPU Scheduling Basic Concepts Scheduling Criteria Scheduling Algorithms Thread Scheduling Multiple-Processor Scheduling Real-Time

More information

Job-Oriented Monitoring of Clusters

Job-Oriented Monitoring of Clusters Job-Oriented Monitoring of Clusters Vijayalaxmi Cigala Dhirajkumar Mahale Monil Shah Sukhada Bhingarkar Abstract There has been a lot of development in the field of clusters and grids. Recently, the use

More information

Ch 4 : CPU scheduling

Ch 4 : CPU scheduling Ch 4 : CPU scheduling It's the basis of multiprogramming operating systems. By switching the CPU among processes, the operating system can make the computer more productive In a single-processor system,

More information

Announcements. Program #1. Program #0. Reading. Is due at 9:00 AM on Thursday. Re-grade requests are due by Monday at 11:59:59 PM.

Announcements. Program #1. Program #0. Reading. Is due at 9:00 AM on Thursday. Re-grade requests are due by Monday at 11:59:59 PM. Program #1 Announcements Is due at 9:00 AM on Thursday Program #0 Re-grade requests are due by Monday at 11:59:59 PM Reading Chapter 6 1 CPU Scheduling Manage CPU to achieve several objectives: maximize

More information

Job sample: SCOPE (VLDBJ, 2012)

Job sample: SCOPE (VLDBJ, 2012) Apollo High level SQL-Like language The job query plan is represented as a DAG Tasks are the basic unit of computation Tasks are grouped in Stages Execution is driven by a scheduler Job sample: SCOPE (VLDBJ,

More information

PROXIMITY AWARE LOAD BALANCING FOR HETEROGENEOUS NODES Mrs. Yogita A. Dalvi Dr. R. Shankar Mr. Atesh Kumar

PROXIMITY AWARE LOAD BALANCING FOR HETEROGENEOUS NODES Mrs. Yogita A. Dalvi Dr. R. Shankar Mr. Atesh Kumar ISSN 2320-9194 1 International Journal of Advance Research, IJOAR.org Volume 1, Issue 9, September 2013, Online: ISSN 2320-9194 PROXIMITY AWARE LOAD BALANCING FOR HETEROGENEOUS NODES Mrs. Yogita A. Dalvi

More information

Grid Computing Systems: A Survey and Taxonomy

Grid Computing Systems: A Survey and Taxonomy Grid Computing Systems: A Survey and Taxonomy Material for this lecture from: A Survey and Taxonomy of Resource Management Systems for Grid Computing Systems, K. Krauter, R. Buyya, M. Maheswaran, CS Technical

More information

Adapting Mixed Workloads to Meet SLOs in Autonomic DBMSs

Adapting Mixed Workloads to Meet SLOs in Autonomic DBMSs Adapting Mixed Workloads to Meet SLOs in Autonomic DBMSs Baoning Niu, Patrick Martin, Wendy Powley School of Computing, Queen s University Kingston, Ontario, Canada, K7L 3N6 {niu martin wendy}@cs.queensu.ca

More information

Chapter 9 Uniprocessor Scheduling

Chapter 9 Uniprocessor Scheduling Operating Systems: Internals and Design Principles, 6/E William Stallings Chapter 9 Uniprocessor Scheduling Patricia Roy Manatee Community College, Venice, FL 2008, Prentice Hall Aim of Scheduling Assign

More information

OPERATING SYSTEMS CS3502 Spring Processor Scheduling. Chapter 5

OPERATING SYSTEMS CS3502 Spring Processor Scheduling. Chapter 5 OPERATING SYSTEMS CS3502 Spring 2018 Processor Scheduling Chapter 5 Goals of Processor Scheduling Scheduling is the sharing of the CPU among the processes in the ready queue The critical activities are:

More information

Operating Systems Unit 3

Operating Systems Unit 3 Unit 3 CPU Scheduling Algorithms Structure 3.1 Introduction Objectives 3.2 Basic Concepts of Scheduling. CPU-I/O Burst Cycle. CPU Scheduler. Preemptive/non preemptive scheduling. Dispatcher Scheduling

More information

On-Line Monitoring of Multi-Area Power Systems in Distributed Environment

On-Line Monitoring of Multi-Area Power Systems in Distributed Environment SERBIAN JOURNAL OF ELECTRICAL ENGINEERING Vol. 3, No. 1, June 2006, 89-101 On-Line Monitoring of Multi-Area Power Systems in Distributed Environment Ramadoss Ramesh 1, Velimuthu Ramachandran 2 Abstract:

More information

CS3733: Operating Systems

CS3733: Operating Systems CS3733: Operating Systems Topics: Process (CPU) Scheduling (SGG 5.1-5.3, 6.7 and web notes) Instructor: Dr. Dakai Zhu 1 Updates and Q&A Homework-02: late submission allowed until Friday!! Submit on Blackboard

More information

Navjot Jyoti ABSTRACT I. INTRODUCTION

Navjot Jyoti ABSTRACT I. INTRODUCTION International Journal of Scientific esearch in Computer Science, Engineering and Information echnology 217 IJSCSEI Volume 2 Issue 1 ISSN : 2456-337 An Analytical eview : Static Load Balancing Algorithms

More information

Co-operative Scheduled Energy Aware Load-Balancing technique for an Efficient Computational Cloud

Co-operative Scheduled Energy Aware Load-Balancing technique for an Efficient Computational Cloud 571 Co-operative Scheduled Energy Aware Load-Balancing technique for an Efficient Computational Cloud T.R.V. Anandharajan 1, Dr. M.A. Bhagyaveni 2 1 Research Scholar, Department of Electronics and Communication,

More information

LOAD BALANCING ALGORITHMS ROUND-ROBIN (RR), LEAST- CONNECTION, AND LEAST LOADED EFFICIENCY

LOAD BALANCING ALGORITHMS ROUND-ROBIN (RR), LEAST- CONNECTION, AND LEAST LOADED EFFICIENCY LOAD BALANCING ALGORITHMS ROUND-ROBIN (RR), LEAST- CONNECTION, AND LEAST LOADED EFFICIENCY Dr. Mustafa ElGili Mustafa Computer Science Department, Community College, Shaqra University, Shaqra, Saudi Arabia,

More information

A New RR Scheduling Approach for Real Time Systems using Fuzzy Logic

A New RR Scheduling Approach for Real Time Systems using Fuzzy Logic Volume 119 No.5, June 2015 A New RR Scheduling Approach for Real Systems using Fuzzy Logic Lipika Datta Assistant Professor, CSE Dept. CEMK,Purba Medinipur West Bengal, India ABSTRACT Round Robin scheduling

More information

Resolving Load Balancing Issue of Grid Computing through Dynamic Approach

Resolving Load Balancing Issue of Grid Computing through Dynamic Approach Resolving Load Balancing Issue of Grid Computing through Dynamic Er. Roma Soni M-Tech Student Dr. Kamal Sharma Prof. & Director of E.C.E. Deptt. EMGOI, Badhauli. Er. Sharad Chauhan Asst. Prof. in C.S.E.

More information

A Comparative Study of Load Balancing Algorithms: A Review Paper

A Comparative Study of Load Balancing Algorithms: A Review Paper Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology ISSN 2320 088X IMPACT FACTOR: 5.258 IJCSMC,

More information

Load Balancing with Random Information Exchanged based Policy

Load Balancing with Random Information Exchanged based Policy Load Balancing with Random Information Exchanged based Policy Taj Alam 1, Zahid Raza 2 School of Computer & Systems Sciences Jawaharlal Nehru University New Delhi, India 1 tajhashimi@gmail.com, 2 zahidraza@mail.jnu.ac.in

More information

Frequently asked questions from the previous class survey

Frequently asked questions from the previous class survey CS 370: OPERATING SYSTEMS [CPU SCHEDULING] Shrideep Pallickara Computer Science Colorado State University L14.1 Frequently asked questions from the previous class survey Turnstiles: Queue for threads blocked

More information

Real-time grid computing for financial applications

Real-time grid computing for financial applications CNR-INFM Democritos and EGRID project E-mail: cozzini@democritos.it Riccardo di Meo, Ezio Corso EGRID project ICTP E-mail: {dimeo,ecorso}@egrid.it We describe the porting of a test case financial application

More information

Distributed Scheduling for the Sombrero Single Address Space Distributed Operating System

Distributed Scheduling for the Sombrero Single Address Space Distributed Operating System Distributed Scheduling for the Sombrero Single Address Space Distributed Operating System Donald S. Miller Department of Computer Science and Engineering Arizona State University Tempe, AZ, USA Alan C.

More information

WebSphere Application Server Base Performance

WebSphere Application Server Base Performance WebSphere Application Server Base Performance ii WebSphere Application Server Base Performance Contents WebSphere Application Server Base Performance............. 1 Introduction to the WebSphere Application

More information

A Resource Discovery Algorithm in Mobile Grid Computing based on IP-paging Scheme

A Resource Discovery Algorithm in Mobile Grid Computing based on IP-paging Scheme A Resource Discovery Algorithm in Mobile Grid Computing based on IP-paging Scheme Yue Zhang, Yunxia Pei To cite this version: Yue Zhang, Yunxia Pei. A Resource Discovery Algorithm in Mobile Grid Computing

More information

Review. Preview. Three Level Scheduler. Scheduler. Process behavior. Effective CPU Scheduler is essential. Process Scheduling

Review. Preview. Three Level Scheduler. Scheduler. Process behavior. Effective CPU Scheduler is essential. Process Scheduling Review Preview Mutual Exclusion Solutions with Busy Waiting Test and Set Lock Priority Inversion problem with busy waiting Mutual Exclusion with Sleep and Wakeup The Producer-Consumer Problem Race Condition

More information

Performance Analysis of Adaptive Dynamic Load Balancing in Grid Environment using GRIDSIM

Performance Analysis of Adaptive Dynamic Load Balancing in Grid Environment using GRIDSIM Performance Analysis of Adaptive Dynamic Load Balancing in Grid Environment using GRIDSIM Pawandeep Kaur, Harshpreet Singh Computer Science & Engineering, Lovely Professional University Phagwara, Punjab,

More information

Operating Systems. Computer Science & Information Technology (CS) Rank under AIR 100

Operating Systems. Computer Science & Information Technology (CS) Rank under AIR 100 GATE- 2016-17 Postal Correspondence 1 Operating Systems Computer Science & Information Technology (CS) 20 Rank under AIR 100 Postal Correspondence Examination Oriented Theory, Practice Set Key concepts,

More information

Scalable Computing: Practice and Experience Volume 10, Number 4, pp

Scalable Computing: Practice and Experience Volume 10, Number 4, pp Scalable Computing: Practice and Experience Volume 10, Number 4, pp. 413 418. http://www.scpe.org ISSN 1895-1767 c 2009 SCPE MULTI-APPLICATION BAG OF JOBS FOR INTERACTIVE AND ON-DEMAND COMPUTING BRANKO

More information

8: Scheduling. Scheduling. Mark Handley

8: Scheduling. Scheduling. Mark Handley 8: Scheduling Mark Handley Scheduling On a multiprocessing system, more than one process may be available to run. The task of deciding which process to run next is called scheduling, and is performed by

More information

Chapter 6: CPU Scheduling. Operating System Concepts 9 th Edition

Chapter 6: CPU Scheduling. Operating System Concepts 9 th Edition Chapter 6: CPU Scheduling Silberschatz, Galvin and Gagne 2013 Chapter 6: CPU Scheduling Basic Concepts Scheduling Criteria Scheduling Algorithms Thread Scheduling Multiple-Processor Scheduling Real-Time

More information

Multiprocessor scheduling

Multiprocessor scheduling Chapter 10 Multiprocessor scheduling When a computer system contains multiple processors, a few new issues arise. Multiprocessor systems can be categorized into the following: Loosely coupled or distributed.

More information

L3.4. Data Management Techniques. Frederic Desprez Benjamin Isnard Johan Montagnat

L3.4. Data Management Techniques. Frederic Desprez Benjamin Isnard Johan Montagnat Grid Workflow Efficient Enactment for Data Intensive Applications L3.4 Data Management Techniques Authors : Eddy Caron Frederic Desprez Benjamin Isnard Johan Montagnat Summary : This document presents

More information

Introduction to Operating Systems Prof. Chester Rebeiro Department of Computer Science and Engineering Indian Institute of Technology, Madras

Introduction to Operating Systems Prof. Chester Rebeiro Department of Computer Science and Engineering Indian Institute of Technology, Madras Introduction to Operating Systems Prof. Chester Rebeiro Department of Computer Science and Engineering Indian Institute of Technology, Madras Week 05 Lecture 18 CPU Scheduling Hello. In this lecture, we

More information

Job Management System Extension To Support SLAAC-1V Reconfigurable Hardware

Job Management System Extension To Support SLAAC-1V Reconfigurable Hardware Job Management System Extension To Support SLAAC-1V Reconfigurable Hardware Mohamed Taher 1, Kris Gaj 2, Tarek El-Ghazawi 1, and Nikitas Alexandridis 1 1 The George Washington University 2 George Mason

More information

Lecture Topics. Announcements. Today: Uniprocessor Scheduling (Stallings, chapter ) Next: Advanced Scheduling (Stallings, chapter

Lecture Topics. Announcements. Today: Uniprocessor Scheduling (Stallings, chapter ) Next: Advanced Scheduling (Stallings, chapter Lecture Topics Today: Uniprocessor Scheduling (Stallings, chapter 9.1-9.3) Next: Advanced Scheduling (Stallings, chapter 10.1-10.4) 1 Announcements Self-Study Exercise #10 Project #8 (due 11/16) Project

More information

A Federated Grid Environment with Replication Services

A Federated Grid Environment with Replication Services A Federated Grid Environment with Replication Services Vivek Khurana, Max Berger & Michael Sobolewski SORCER Research Group, Texas Tech University Grids can be classified as computational grids, access

More information

CSE120 Principles of Operating Systems. Prof Yuanyuan (YY) Zhou Scheduling

CSE120 Principles of Operating Systems. Prof Yuanyuan (YY) Zhou Scheduling CSE120 Principles of Operating Systems Prof Yuanyuan (YY) Zhou Scheduling Announcement l Homework 2 due on October 26th l Project 1 due on October 27th 2 Scheduling Overview l In discussing process management

More information

Advanced Topics UNIT 2 PERFORMANCE EVALUATIONS

Advanced Topics UNIT 2 PERFORMANCE EVALUATIONS Advanced Topics UNIT 2 PERFORMANCE EVALUATIONS Structure Page Nos. 2.0 Introduction 4 2. Objectives 5 2.2 Metrics for Performance Evaluation 5 2.2. Running Time 2.2.2 Speed Up 2.2.3 Efficiency 2.3 Factors

More information

International Journal of Advanced Research in Computer Science and Software Engineering

International Journal of Advanced Research in Computer Science and Software Engineering Volume 2, Issue 11, November 2012 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Process Scheduling

More information

Introduction to Grid Computing

Introduction to Grid Computing Milestone 2 Include the names of the papers You only have a page be selective about what you include Be specific; summarize the authors contributions, not just what the paper is about. You might be able

More information

DISTRIBUTED HIGH-SPEED COMPUTING OF MULTIMEDIA DATA

DISTRIBUTED HIGH-SPEED COMPUTING OF MULTIMEDIA DATA DISTRIBUTED HIGH-SPEED COMPUTING OF MULTIMEDIA DATA M. GAUS, G. R. JOUBERT, O. KAO, S. RIEDEL AND S. STAPEL Technical University of Clausthal, Department of Computer Science Julius-Albert-Str. 4, 38678

More information

Performance Characterization of the Dell Flexible Computing On-Demand Desktop Streaming Solution

Performance Characterization of the Dell Flexible Computing On-Demand Desktop Streaming Solution Performance Characterization of the Dell Flexible Computing On-Demand Desktop Streaming Solution Product Group Dell White Paper February 28 Contents Contents Introduction... 3 Solution Components... 4

More information

Study of Load Balancing Schemes over a Video on Demand System

Study of Load Balancing Schemes over a Video on Demand System Study of Load Balancing Schemes over a Video on Demand System Priyank Singhal Ashish Chhabria Nupur Bansal Nataasha Raul Research Scholar, Computer Department Abstract: Load balancing algorithms on Video

More information

CPU Scheduling Algorithms

CPU Scheduling Algorithms CPU Scheduling Algorithms Notice: The slides for this lecture have been largely based on those accompanying the textbook Operating Systems Concepts with Java, by Silberschatz, Galvin, and Gagne (2007).

More information

Simulation Analysis of Linear Programming Based Load Balancing Algorithms for Routers

Simulation Analysis of Linear Programming Based Load Balancing Algorithms for Routers Simulation Analysis of Linear Programming Based Load Balancing Algorithms for Routers School of Computer Science & IT Devi Ahilya University, Indore ABSTRACT The work in this paper is the extension of

More information

Performance and Scalability with Griddable.io

Performance and Scalability with Griddable.io Performance and Scalability with Griddable.io Executive summary Griddable.io is an industry-leading timeline-consistent synchronized data integration grid across a range of source and target data systems.

More information

QoS Guided Min-Mean Task Scheduling Algorithm for Scheduling Dr.G.K.Kamalam

QoS Guided Min-Mean Task Scheduling Algorithm for Scheduling Dr.G.K.Kamalam International Journal of Computer Communication and Information System(IJJCCIS) Vol 7. No.1 215 Pp. 1-7 gopalax Journals, Singapore available at : www.ijcns.com ISSN: 976 1349 ---------------------------------------------------------------------------------------------------------------------

More information

Design and Evaluation of I/O Strategies for Parallel Pipelined STAP Applications

Design and Evaluation of I/O Strategies for Parallel Pipelined STAP Applications Design and Evaluation of I/O Strategies for Parallel Pipelined STAP Applications Wei-keng Liao Alok Choudhary ECE Department Northwestern University Evanston, IL Donald Weiner Pramod Varshney EECS Department

More information

Load Balancing Algorithm over a Distributed Cloud Network

Load Balancing Algorithm over a Distributed Cloud Network Load Balancing Algorithm over a Distributed Cloud Network Priyank Singhal Student, Computer Department Sumiran Shah Student, Computer Department Pranit Kalantri Student, Electronics Department Abstract

More information

CHAPTER 6 STATISTICAL MODELING OF REAL WORLD CLOUD ENVIRONMENT FOR RELIABILITY AND ITS EFFECT ON ENERGY AND PERFORMANCE

CHAPTER 6 STATISTICAL MODELING OF REAL WORLD CLOUD ENVIRONMENT FOR RELIABILITY AND ITS EFFECT ON ENERGY AND PERFORMANCE 143 CHAPTER 6 STATISTICAL MODELING OF REAL WORLD CLOUD ENVIRONMENT FOR RELIABILITY AND ITS EFFECT ON ENERGY AND PERFORMANCE 6.1 INTRODUCTION This chapter mainly focuses on how to handle the inherent unreliability

More information

Adaptive Cluster Computing using JavaSpaces

Adaptive Cluster Computing using JavaSpaces Adaptive Cluster Computing using JavaSpaces Jyoti Batheja and Manish Parashar The Applied Software Systems Lab. ECE Department, Rutgers University Outline Background Introduction Related Work Summary of

More information

A Load Balancing Approach to Minimize the Resource Wastage in Cloud Computing

A Load Balancing Approach to Minimize the Resource Wastage in Cloud Computing A Load Balancing Approach to Minimize the Resource Wastage in Cloud Computing Sachin Soni 1, Praveen Yadav 2 Department of Computer Science, Oriental Institute of Science and Technology, Bhopal, India

More information

GRID SIMULATION FOR DYNAMIC LOAD BALANCING

GRID SIMULATION FOR DYNAMIC LOAD BALANCING GRID SIMULATION FOR DYNAMIC LOAD BALANCING Kapil B. Morey 1, Prof. A. S. Kapse 2, Prof. Y. B. Jadhao 3 1 Research Scholar, Computer Engineering Dept., Padm. Dr. V. B. Kolte College of Engineering, Malkapur,

More information

Frequently asked questions from the previous class survey

Frequently asked questions from the previous class survey CS 370: OPERATING SYSTEMS [CPU SCHEDULING] Shrideep Pallickara Computer Science Colorado State University L15.1 Frequently asked questions from the previous class survey Could we record burst times in

More information

MOHA: Many-Task Computing Framework on Hadoop

MOHA: Many-Task Computing Framework on Hadoop Apache: Big Data North America 2017 @ Miami MOHA: Many-Task Computing Framework on Hadoop Soonwook Hwang Korea Institute of Science and Technology Information May 18, 2017 Table of Contents Introduction

More information

Analysis of Various Load Balancing Techniques in Cloud Computing: A Review

Analysis of Various Load Balancing Techniques in Cloud Computing: A Review Analysis of Various Load Balancing Techniques in Cloud Computing: A Review Jyoti Rathore Research Scholar Computer Science & Engineering, Suresh Gyan Vihar University, Jaipur Email: Jyoti.rathore131@gmail.com

More information

CPU Scheduling. Daniel Mosse. (Most slides are from Sherif Khattab and Silberschatz, Galvin and Gagne 2013)

CPU Scheduling. Daniel Mosse. (Most slides are from Sherif Khattab and Silberschatz, Galvin and Gagne 2013) CPU Scheduling Daniel Mosse (Most slides are from Sherif Khattab and Silberschatz, Galvin and Gagne 2013) Basic Concepts Maximum CPU utilization obtained with multiprogramming CPU I/O Burst Cycle Process

More information

Unit 3 : Process Management

Unit 3 : Process Management Unit : Process Management Processes are the most widely used units of computation in programming and systems, although object and threads are becoming more prominent in contemporary systems. Process management

More information

Performance Enhancement of Scheduling Algorithms in Web Server Clusters using Improved Dynamic Load Balancing Policies

Performance Enhancement of Scheduling Algorithms in Web Server Clusters using Improved Dynamic Load Balancing Policies Computing For Nation Development, February 08 09, 2008 Bharati Vidyapeeth s Institute of Computer Applications and Management, New Delhi Performance Enhancement of Scheduling Algorithms in Web Server Clusters

More information

Last Class: Processes

Last Class: Processes Last Class: Processes A process is the unit of execution. Processes are represented as Process Control Blocks in the OS PCBs contain process state, scheduling and memory management information, etc A process

More information

Uniprocessor Scheduling. Chapter 9

Uniprocessor Scheduling. Chapter 9 Uniprocessor Scheduling Chapter 9 1 Aim of Scheduling Assign processes to be executed by the processor(s) Response time Throughput Processor efficiency 2 3 4 Long-Term Scheduling Determines which programs

More information

Available online at ScienceDirect. Procedia Computer Science 93 (2016 )

Available online at   ScienceDirect. Procedia Computer Science 93 (2016 ) Available online at www.sciencedirect.com ScienceDirect Procedia Computer Science 93 (2016 ) 269 275 6th International Conference On Advances In Computing & Communications, ICACC 2016, 6-8 September 2016,

More information

ayaz ali Micro & Macro Scheduling Techniques Ayaz Ali Department of Computer Science University of Houston Houston, TX

ayaz ali Micro & Macro Scheduling Techniques Ayaz Ali Department of Computer Science University of Houston Houston, TX ayaz ali Micro & Macro Scheduling Techniques Ayaz Ali Department of Computer Science University of Houston Houston, TX 77004 ayaz@cs.uh.edu 1. INTRODUCTION Scheduling techniques has historically been one

More information

A Study of the Performance Tradeoffs of a Tape Archive

A Study of the Performance Tradeoffs of a Tape Archive A Study of the Performance Tradeoffs of a Tape Archive Jason Xie (jasonxie@cs.wisc.edu) Naveen Prakash (naveen@cs.wisc.edu) Vishal Kathuria (vishal@cs.wisc.edu) Computer Sciences Department University

More information

Module 5 Introduction to Parallel Processing Systems

Module 5 Introduction to Parallel Processing Systems Module 5 Introduction to Parallel Processing Systems 1. What is the difference between pipelining and parallelism? In general, parallelism is simply multiple operations being done at the same time.this

More information

Computer Hardware and System Software Concepts

Computer Hardware and System Software Concepts Computer Hardware and System Software Concepts Introduction to concepts of Operating System (Process & File Management) Welcome to this course on Computer Hardware and System Software Concepts 1 RoadMap

More information

SMD149 - Operating Systems - Multiprocessing

SMD149 - Operating Systems - Multiprocessing SMD149 - Operating Systems - Multiprocessing Roland Parviainen December 1, 2005 1 / 55 Overview Introduction Multiprocessor systems Multiprocessor, operating system and memory organizations 2 / 55 Introduction

More information

Overview. SMD149 - Operating Systems - Multiprocessing. Multiprocessing architecture. Introduction SISD. Flynn s taxonomy

Overview. SMD149 - Operating Systems - Multiprocessing. Multiprocessing architecture. Introduction SISD. Flynn s taxonomy Overview SMD149 - Operating Systems - Multiprocessing Roland Parviainen Multiprocessor systems Multiprocessor, operating system and memory organizations December 1, 2005 1/55 2/55 Multiprocessor system

More information

Enhancing the Performance of Feedback Scheduling

Enhancing the Performance of Feedback Scheduling Enhancing the Performance of Feedback Scheduling Ayan Bhunia Student, M. Tech. CSED, MNNIT Allahabad- 211004 (India) ABSTRACT Feedback scheduling is a kind of process scheduling mechanism where process

More information