VM Management for Cross-Cloud Computing Environment

Transcription

1 2012 International Conference on Communication Systems and Network Technologies VM Management for Cross-Cloud Computing Environment Deepti Theng M. Tech. Computer Science and Engineering G. H. Raisoni College of Engineering, Nagpur, Maharashtra, India K. N. Hande Assistant Professor, Department of CSE, G. H. Raisoni College of Engineering, Nagpur, Maharashtra, India Abstract Cloud computing is the set of distributed computing nodes. The distribution of virtual machine (VM) images to a set of distributed compute nodes in a Cross- Cloud computing environment is main issue considered in this paper. This paper will be dealing with the problem of scheduling virtual machine (VM) images to a set of distributed compute nodes in a Cross-Cloud computing environment. To address this problem, an efficient approach for VM management is proposed and implemented in this paper. The result will be used as an effective scheduling guidance for VM scheduling on cloud computing as well as cross cloud computing environment. Keywords: Virtualization; cloud computing; virtual machine (VM) image; cross-cloud computing; scheduling I. INTRODUCTION Distributed computing has achieved high growth, due its feature of resource sharing and providing user a cost effective high computational speed. Cloud computing and virtualization are two of the major IT trends of the decade. There are so many areas of implications provided by cloud computing which includes, enhanced data computation, cost-effective, decrement in operational resources, optimized resource deployment, enhanced software development, mobility in data access, security, etc [1]. Thus to provide cost effective computations, cloud environment needs to be operated in an efficient way. After virtualization it has been possible to present compute resources in the form of Virtual Machine (VM) Images. Virtualization plays a special role in cloud computing. Typically VMs are offered in different types, each type have its own characteristics which includes number of CPU cores, amount of main memory, etc. and cost. To satisfy the needs of such a high performance computations evolves the use of federated clouds and multi-cloud deployments [2], [3], [4]. A Cross-Cloud computing environment is the interaction of two or more Cloud computing sites offering different computations. If the Cloud paradigm is to be extended to facilitate cross provider utilization of resources, several challenges need to be solved [5]. One of the requirements for Cross-Cloud computing discussed in this paper is the cost effective and secured scheduling of virtual machine images. The deployment mechanism should be able to cope with both cluster environments and more heterogeneous, distributed Cross- Cloud computing environments, which also includes desktop systems as found in private Clouds. The main purpose of scheduling here is to find a best match of compute resources with the task the job consist of. This paper is organized as follows. First the literature review on self-scheduling algorithm is given in section 2. Then new proposed extension of FSS that is, HFSS elaborated in section 3 as well as proposed system architecture is discussed. Then section 4 concludes the paper giving idea of the future work. II. EXISTING SCHEDULING SCHEMES This section describes different scheduling algorithms best suitable for the distributed and highly heterogeneous environment for VM image scheduling. A. Loop Scheduling Schemes In static even scheduling loop iteration space is divided into n chunks evenly and then each of these chunks is assigned to a processor. When a chunk of loop iteration is dispatched at runtime by an idle processor is called dynamic scheduling. Guided scheduling is somewhat similar to dynamic scheduling, but chunk size keep shrinking per dispatch. Cross-Cloud computing platform is distributed, dynamically changing and extremely heterogeneous, which may contain clouds, cloud of clouds, etc. Considering requirements of such an environment loop scheduling schemes are seems to be suitable for effective scheduling. Basically loop scheduling schemes are classified according to the time at which scheduling decision is made. According to the author of [6], dynamic loop scheduling schemes are more appropriate for load balancing in the heterogeneous cross-cloud computing environment /12 $ IEEE DOI /CSNT

2 B. Self-Scheduling Schemes Efficient scheduling on distributed systems such as heterogeneous cluster and grid computing is mostly affected by load balancing. This section presents a general approach toward different self scheduling schemes. Several best performing self scheduling schemes are reviewed here as below: Guided Self-Scheduling (GSS): It is characterized by its scheduling strategy. It can dynamically change the number of iterations assigned to each processor [7]. In this chunk size is dynamically calculated as It sometimes offer good load balance with small scheduling overhead, by assigning large size chunks at the beginning of a parallel loop. Factorizing Self-Scheduling (FSS): Factorizing selfscheduling results from statistical analysis. Tasks are scheduled in batches of P chunks of equal size as Only a subset of the remaining tasks is scheduled in each batch. FSS offers better load balance as compared to the GSS when execution time changes widely and randomly [8]. Trapezoid Self-Scheduling (TSS): It maintains reasonable load balance while trying to reduce the need for synchronization [9]. It linearly decreases the chunk size. Large chunks of iterations are allocated to the first few processors and successively smaller chunks to the last few processors. In [8], [9] the author has proposed new improved selfscheduling schemes named NGSS and ANGSS. NGSS was originally derived from GSS plus self-scheduling scheme. A self-scheduling divides the workload in to two partitions; first % workload will be dispatched according to their performance while remaining (100- )% workload dispatched according to well defined selfscheduling scheme. Next the scheme is enhanced from GSS to ANGSS, in which the types of loop iterations were classified into four classes. In [6], the author has shown implementation of these scheduling schemes for the purpose of evaluation. According to the results obtained by [6], the author says that traditional self scheduling schemes (FSS and TSS) perform slightly better than NGSS and ANGSS. From this discussion FSS scheme is selected for implementation with some new innovations added to achieve efficiency in extremely heterogeneous environment. This improved FSS is termed as HFSS, were H means Heterogeneous here. III. RELATED WORK In this section first HFSS is defined. Then, the Cross- Cloud platform with proposed system architecture is presented. A. Heterogeneous Factorizing Self-Scheduling Scheme Based on the literature review outlined in the previous section new FSS i.e., HFSS is proposed here. HFSS takes many ideas and characteristics of previous scheduling framework to provide best match to the extremely heterogeneous and geologically distributed Cross-Cloud environment. Factorizing self-scheduling is said to be heterogeneous as it would offer Better load balance when execution time changes widely and randomly. Dynamic adoption for execution of the application Partitioning of loop iterations according to the current performance The modified and proposed algorithm is as below: At Scheduler: Accept VM request of user at scheduler Scheduler will detect corresponding servers Partition the % of the load to one computing node (which is minimum loaded) Remaining (100- ) % of the workload will be dispatched on other computing nodes giving good load balance Computational result will be accepted from corresponding computing node and provided to the user The process continues for the each coming request At Compute Node: Compute node assigned for the task will track its execution time Finally the time required for each task will be calculated and result is evaluated This partitioning provides better load balance and will boost the performance of cloud. Working example is explained in the next subsection of system architecture. B. Proposed System Architecture The scheduler will be implemented as shown in figure 1. In this the L-Scheduler for multi-cloud environment will execute the scheduling algorithm mentioned above. The scenario is best depicted as below

3 First it goes to the respective serve requested for computation. If the request is given to our cloud then it is scheduled to the best computational node (actual compute nodes) with the help of proposed algorithm by the L-Scheduler shown in Fig. 1. Once the request reaches the desired server, then the client-server gets connected and computations are carried out. This process goes on continuously for all the incoming requests. Fig. 1 Cross-Cloud Scheduler Deployment Cloud Server-1: It will act as a Server for Student Management. In present scenario, managing students in educational institute is a bigger challenge. This will provide a solution to the day to day administrative operations and to carry out educational institute functions smartly. This server has two clouds providing same functionalities. The environment created here is termed as multi-cloud environment. The server will select cloud with minimum load. Cloud Server-2: It will provide facility for Finance Management. This site will provide customers to manage banking from anywhere. Such banking solutions have many features including transactional as well as non-transactional. It also provides financial institution administration, management of multiple users, personal financial management support. This workflow is followed smoothly with the peer-topeer computing network. This network provides high scalability, as it is possible to add more peers to the system and can scaled to larger network [11]. Thus it is good for large scale computations. On one cloud server of student management, the two servers will hold two different databases for students. The incoming request will be partitioned and will be getting computed on the respective servers holding corresponding required database. In this way request will be partitioned and effective load balance at heterogeneous environment will be achieved. System will be evaluated by such a real life applications at each cloud server. C. Experimental Results This section shows some experimental results Clod Server-3: Matrix Multiplication. On this server large matrix multiplication will be performed. This will be executed for different large matrix multiplication, to check response time and computational speed. These services will be designed at each cloud server which provide above mentioned functionality at the respective cloud server Work Flow: In this detailed execution plan is given for a job. Whenever a request comes from a client it will be first accepted at the appropriate cloud server. At this local cloud site the incoming request will be served by the scheduler to better match the requirement at that cloud server. This is elaborated in the following steps: The request comes from a client. Fig. 2 Cloud Server-1 Showing Student Management

4 work is in progress, subsequent paper will include the complete result of proposed algorithm showing effective VM scheduling. This work will give a novel idea of crosscloud VM scheduling. Fig. 3 Database for Student Management at Computational Node-1 Fig. 4 Database for Student Management at Computational Node-2 These are some sample results for one cloud server. Others are under working process. The final result will explain the brief execution of algorithm. While testing the computational time will be measured for different VM request. The result of same will be made available in the next version of this paper. When multiple requests are sent by the different users, then first few requests will be assigned to the minimally loaded servers. Remaining requests are scheduled on rest of the computing nodes. IV. CONCLUSION In this paper an existing loop scheduling algorithms are presented. From previous study a Heterogeneous Factorizing Self-Scheduling algorithm is proposed and considered for the implementation for virtual machine scheduling in cross-cloud environment. Such an implementation is not yet done with cross-cloud platform which is proposed here. System architecture is proposed with detailed execution plan on each cloud server. The References [1] Ammar Khalid, Husnain Mujtaba, Data Processing Issues in Cloud Computing, Second International Conference on Machine Vision, [2] R. Moreno-Vozmediano, R. S. Montero, I. M. Llorente, Multicloud Deployment of Computing Clusters for Loosely Coupled MTC Applications, IEEE Transaction on Parallel and Distributed Systems, Vol 22, no 6, pp , June, [3] B. Rochwerger, D. Beritgand, E. Levy, A. Galis and others, The Reservoir Model and Architecture for Open Federated Cloud Computing, IBM Journal of Rsearch and Development, Vol 53, no 4, pp. 1-11, July, [4] Jose Luis Lucas Simarro, Rafael Moreno-Vozmediano, Ruben S. Montero and I. M. Llorente, Dynamic placement of virtual machines for cost optimization in multi-cloud environment, International Conference on High Performance Computing and Simulation (HPCS), 2011, pp [5] Matthias Schmidt, Niels Fallenbeck, Matthew Smith, Bernd Freisleben, Efficient Distribution of Virtual Machines for Cloud Computing, 18th Euromicro Conference on Parallel, Distributed and Network-based Processing, [6] Wen-Chung Shih, Shian-shyong Tseng, Performance Study of Parallel Programming on Cloud Computing Environment: Using MapReduce, IEEE Computer Sociaty, [7] Polychronopoulos, C.D. and D.J. Kuck, Guided self-scheduling: A practical scheduling scheme for parallel supercomputers, 1987, IEEE Computer Society. p [8] Yang, C.T., K.W. Cheng, and K.C. Li, An Efficient Parallel Loop Self- Scheduling Scheme for Cluster Environments, Journal of Supercomputing, : p [9] Yang, C.T. and S.C. Chang, A Parallel Loop Self-Scheduling on Extremely Heterogeneous PC Clusters, Journal of Information Science and Engineering, (2): p [10] Wen-Chung Shih, Chao-Tung Yang*, Ping-I Chen, Shian-Shyong Tseng, A Hybrid Parallel Loop Scheduling Scheme on Heterogeneous PC Clusters, Sixth International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT 2005, pp [11] Qi CAO Satoshi FUJITA, Load Balancing Schemes for a Hierarchical Peer-to-Peer File Search System, International Conference on P2P, Parallel, Grid, Cloud and Internet Computing, [12] Sharrukh Zaman, Daniel Grosu, A Distributed Algorithm for the Replica Placement Problem, IEEE Transactions on Parallel and Distributed Systems, [13] Hai Zhong, Kun Tao, Xuejie Zhang, An Approach to Optimized Resource Scheduling Algorithm for Open-Source Cloud Systems, Fifth Annual ChinaGrid Conference (ChinaGrid), 2010, pp, [14] Sumit Kumar Bose, Scott Brock, Ronald Skeoch, Shrisha Rao, CloudSpider: Combining Replication with Scheduling for Optimizing Live Migration of Virtual Machines across Wide Area Networks, 11th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGrid), 2011, pp [15] Daniel Warneke, Exploiting Dynamic Resource Allocation for Efficient Parallel Data Processing in the Cloud, IEEE Transactions on Parallel and Distributed Systems, Vol. 22, No. 6, June

5 [16] Shu-Ching, Kuo-Qin, Shun-Sheng, Ching-Wei, A Three-Phases Scheduling in a Hierarchical Cloud Computing Network, Third International Conference on Communications and Mobile Computing, [17] Andrew J. Younge, Gregor von Laszewski, Lizhe Wang, Efficient Resource Management for Cloud Computing Environments, 10th IEEE International Conference on Cloud and Grid Computing, 2010 [18] R. Jeyarani, N. Nagaveni, R. Vasanth Ram, Design and Implementation of an efficient Two-level Scheduler for Cloud Computing Environment, 10th IEEE/ACM International Conference on Cluster, Cloud and Grid Computing, 2010 [19] Lizhe Wang, Gregor von Laszewski, Marcel Kunze, Jie Tao, Schedule Distributed Virtual Machines in a Service Oriented Environment, 24th IEEE International Conference on Advanced Information Networking and Applications,