INTERNATIONAL JOURNAL OF TECHNOLOGY AND COMPUTING (IJTC) ISSN X, Volume 2, Issue 7 July 2016.

Transcription

1 A REVIEW ON BALANCING THE LOAD ON CLOUD USING ACCLB HYBRID LOAD BALANCING TECHNIQUE A B S T R A C T Gagandeep Kaur*, Er. Sushil Kamboj SUSGOI, Tangori Gagandeepkaur1087@gmail.com*, er.kamboj@gmail.com Cloud Computing is an essential ingredient of modern computing systems. Computing concepts, technology and architectures have been developed and consolidated in the last decades; many aspects are subject to technological evolution and revolution. Cloud Computing is an computing technology that is rapidly consolidating itself as the next step in the development and deployment of increasing number of distributed application. The main objective of load balancing methods is to speed up the execution of applications on resources whose workload varies at run time in unpredictable way. Load balancing techniques are widely discussed in homogeneous as well as heterogeneous environments such as grids. There are basically two kinds of load balancing techniques. They are (i) Static and (ii) dynamic. In cloud computing environments, whenever a VM is heavily loaded with multiple tasks, these tasks have to be removed and submitted to the under loaded VMs of the same data center. In this case, when we remove more than one task from a heavy loaded VM and if there is more than one VM available to process these tasks, the tasks have to be submitted to the VM such that there will be a good mix of priorities i.e., no task should wait for a long time in order to get processed. Load balancing is done at virtual machine level i.e., at intra-data center level. We are proposing the ACCLB to balance the load on the cloud and compared it with the existing load balancing methods such as Vector Dot, Join idle queue. The main objective will be to balance the load on cloud and to reduce the energy consumption as compared to previous, on the cloud by using proposed method. Also we have to prove that our proposed technique is more efficient for load balancing and energy consumption on cloud as compared to previous. Keywords: ACCLB, Join idle queue, Load balancing, Cloudsim, IASS, PASS, SAAS I. INTRODUCTION Cloud computing is a distributed computing paradigm in which a pool of computing resources such as virtualized physical machines which host applications, shared storage devices like NFS (Network File Storage), backup servers etc. are available to users via Internet. Prevalent use of cloud computing has resulted advancement in the number of hosting datacenters which have brought forth many concerns, including the cost IJTC www. ijtc.org 347

2 of electrical energy, cooling, peak power dissipation and carbon emission. As the workload on the applications differs from time to time, it results in different resource requirements and hence dynamic efficient use of these shared resources is one of the crucial technical problems. The issue of tackling high energy use can be addressed by removing improficiencies and waste which occurs in the way computing resources get involved to serve application workloads and also how the electricity is carried to computing resources,. This can be done by improving both the physical infrastructure of data centers, and the resource allocation and management algorithms. In cloud model when a node leaves the cloud, the application running on it will be transfer to another node [1, 11]. A load balancing strategy is required to maintain the load among all computing nodes. Load balancing strategy is of two types, one is static and second is dynamic load balancing [10, 12]. Making a decision for Load balancing is very important task because it may affect the overall performance. Therefore decision is made by considering the parameters like network, application characteristics and computing node power [8]. Load balancing decision may automated or manual. A dynamic load balancing strategy is best suited in cloud environment due to its dynamic nature. The overall performance may get affected by different parameter together [13]. If a single parameter is considered for decision making, the performance may limit because other parameters also affect the performance. E.g. only available bandwidth is considered for load balancing and applications are transfer to another node, but computing power of that node may be very low. In that condition there will be a large queue on the node and the overall performance may limit [14]. A. CLOUD COMPUTING APPLICATION ARCHITECTURE As we know that cloud computing is the shift of computing to a host of hardware infrastructure that is distributed in the cloud. The commodity hardware infrastructure consists of the various low cost data servers that are connected to the system and provide their storage and processing and other computing resources to the application. Cloud computing involves running applications on virtual servers that are allocated on this distributed hardware infrastructure available in the cloud. These virtual servers are made in such a way that the different service level agreements and reliability issues are met. There may be multiple instances of the same virtual server accessing the different parts of the hardware infrastructure available. This is to make sure that there are multiple copies of the applications which are ready to take over on another one s failure. The virtual server distributes the processing between the infrastructure and the computing is done and the result returned. There will be a workload distribution management system, also known as the grid engine, for managing the different requests coming to the virtual servers. This engine will take care of the creation of multiple copies and also the preservation of integrity of the data that is stored in the infrastructure. This will also adjust itself such that even on heavier load, the processing is completed as per the requirements. The different workload management systems are hidden from the users. For the user, the processing is done and the result is obtained. IJTC www. ijtc.org 348

3 There is no question of where it was done and how it was done. The users are billed based on the usage of the system - as said before - the commodity is now cycles and bytes. The billing is usually on the basis of usage per CPU per hour or GB data transfer per hour. Fig 1. Cloud Computing Application Architecture B. LOAD BALANCING ON CLOUD COMPUTING Load balancing is usually applied on huge amount of data traffic and servers to distribute work. Advanced architectures in cloud are adopted to achieve speed and efficiency. There are several characteristics of load balancing such as: equal division of work across all the nodes, facilitation in achieving user satisfaction, improve overall performance of system, reduce response time, and provide services to achieve complete resource utilization [4]. As an example, if we make one application on cloud and hundreds of users are expected to access it at any one time. Therefore, response time to hundred people will be very slow and servers will become busy very quickly, resulting in slow response and unsatisfactory users. If we apply load balancing on our application, then work will be distributed at other nodes and we can get high performance and better response [5]. Load balancing of large distributed server systems is a complex optimization problem of critical importance in cloud systems and data centers. Existing schedulers often incur a high communication overhead when collecting the data required to make scheduling decisions, hence delaying job requests on their way to the executing servers. We propose a novel scheme that incurs no communication overhead between the users and the servers upon job arrival, thus removing any scheduling overhead from the job s critical path. Parameters for Load Balancing There are mainly three parameters and their sub-parameters which may participate in the decision process of load balancing and hence may affect the performance of cloud system. Network parameter has some subparameters like Inter-communication delay, available bandwidth, and communication link power and network latency. No. of available Parallel Elements, processing power and memory capacity are the characteristics of IJTC www. ijtc.org 349

4 computing node. Application can affect the performance due to its execution time, pre-emptive and non-preemptive characteristics. C. POLICIES OR STRATEGIES IN DYNAMIC LOAD BALANCING There are 4 policies [4]: Transfer Policy: The part of the dynamic load balancing algorithm which selects a job for transferring from a local node to a remote node is referred to as Transfer policy or Transfer strategy. Selection Policy: It specifies the processors involved in the load exchange (processor matching) Location Policy: The part of the load balancing algorithm which selects a destination node for a transferred task is referred to as location policy or Location strategy. Information Policy: The part of the dynamic load balancing algorithm responsible for collecting information about the nodes in the system is referred to as Information policy or Information strategy. II. RELATED WORK In complex and large systems, there is a tremendous need for energy efficiency as well as load balancing on cloud environment. For simplifying such needs globally one thing which can be done is, employing techniques would act at the components of the clouds in such a way that the load of the whole cloud is balanced and results in energy efficiency. Sidhu et al. (2016) [1] proposed algorithm Vector Dot gives the better results as compared to Active clustering and Join Idle queue. We analyzed the results on the basis of different performance parameters such as Response time, Execution time and Energy consumption. Cloud computing has recently become popular due to the maturity of related technologies such as network devices, software applications and hardware capacities. Resources in these systems can be widely distributed and the scale of resources involved can range from several servers to an entire data center. In Cloud Computing Load balancing has a great role. To allocate and balance the load of the resources among the various components and nodes load balancing is required. Ebin Deni Raj et al. (2015) [2] described Big Data and parallel computing are used extensively for processing large quantities of data, structured, semi structured or totally unstructured. MapReduce and Hadoop are used for the parallel data processing of these kinds of data. Various scheduling policies are used for MapReduce scheduling which is discussed in detail and a new scheduling technique Two Phase Scheduling Policy (TPSP) based resource allocation for MapReduce is implemented and the efficiency is verified. Wei Deng et al. (2014) [3] proposed the Harnessing Renewable Energy in Cloud Datacenters. They provided taxonomy of the state-of-the-art research in applying renewable energy in cloud computing datacenters from 5 key aspects, including propagation models and prediction methods of renewable energy, content planning of IJTC www. ijtc.org 350

5 green datacenters, intra-datacenter workload scheduling and load balancing across geographically distributed datacenters. Ching-Hsien Hsu et al. (2014) [4] discussed Optimizing energy consumption with task consolidation in clouds. They presented Energy aware Task Consolidation (ETC) technique to minimize energy consumption. Taking into consideration the architecture of most cloud systems, a default CPU utilization threshold of 70% is used to demonstrate task consolidation management amongst virtual clusters. The simulation result shows that ETC can significantly reduce power consumption when managing task consolidation for cloud systems. ETC has up to 17% improvement over a recent work that reduces energy consumption by maximizing resource utilization. Siva Theja Maguluri et al. (2014) [5] discussed a stochastic model of cloud computing, where jobs appear according to a random process and request resources like CPU, memory and storage space. They considered a model where the resource allocation problem can be separated into a load balancing problem and a scheduling problem. They studied the performance of join-the-shortest-queue (JSQ) routing and MaxWeight scheduling policy under this model. The result shows that it is heavy traffic optimal when all the servers are identical and also found that using the power-of-two-choices routing instead of JSQ routing is also heavy traffic optimal. Then they considered a simpler setting where the jobs are of the same type, therefore only load balancing is needed. It has been established by others using diffusion limit arguments that the power-of-two-choices algorithm is heavy traffic optimal. Young Myoung Ko et al. (2014) [6] proposed a distributed speed scaling and load balancing algorithm considering the tail probability of virtual waiting times. They presented a novel distributed algorithm that adjusts server speed and workload routing according to the demand of multiple (virtualized) applications. Unlike former studies that use the mean response time of a simplem/m/1 queue, the proposed algorithm incorporates the response time distribution obtained from a G/G/1/PS queue. They explain the optimality of the proposed algorithm and discussed the impact of incorporating response time constraints. The numerical result shows the effectiveness of the algorithm when implemented in a distributed manner. Huangke Chen et al. (2014) [7] proposed an uncertainty-aware scheduling architecture for a cloud data center, and developed a novel scheduling algorithm, namely PRS, to make good trade-offs among tasks guaranteeing ratio, system s resource utilization, system s energy consumption and stability. To improve energy efficiency, they proposed three strategies to scale up and down the system s computing resources according to workload to improve resource utilization and to reduce energy consumption for the cloud data center. To evaluate the effectiveness of PRS, they conducted extensive simulation experiments with both the synthetic workloads and Google workload traces. Experimental results showed the effectiveness of the algorithm PRS compared with other related algorithms. IJTC www. ijtc.org 351

6 III. RESEARCH PROBLEM In a cloud environment, there may be any number of host machines and each host machine has differentdifferent load due to virtual machines as per the client s demand. The load of a host machine may be of various types such as CPU load, Memory load, Storage load and Network related load etc. If the load of any host machine exceeds its capacity then it affects its efficiency. In runtime, any client application service may change their resource (CPU, RAM, Storage and Bandwidth etc.) demand and this causes the host system to be imbalanced. If this imbalanced situation occurs due to overloading then system is balanced using load balancing techniques by distributing the extra workload to the whole clouds host heaving light loads. This helps to improve the overall performance of the cloud system. Load balancing can be define as a method for distributing workload on the multiple computers or a computer cluster through network links to achieve optimal resource utilization which maximizes throughput and minimizes overall response time. It minimizes the total waiting time of the resources as well as avoids too much overload on the resources. In this technique traffic is divided among servers, so that data can be sent and received without maximum delay. Load balancing is a methodology to distribute work load across multiple computers, or other resources over the network links to achieve the optimal resource utilization, minimum data processing, minimum average response time and avoid overhead. In the past number of load balancing algorithms have been developed specifically to suit the dynamic cloud computing environments such as WLC (Weighted Least Connection) algorithm, LBMM (Load Balancing Min- Min) algorithm, Bee-MMT (Artificial Bee Colony algorithm Minimal Migration Time), active Clustering algorithm, Honeybee Foraging Algorithm. In paper, we are proposing the ACCLB- based on ant colony and complex network theory (ACCLB) to balance the load on the cloud and compared it with the existing load balancing methods such as Vector Dot and Join idle queue. The main objective of the research is balance the load on cloud and consumes less energy as compared to previous, on the cloud by using proposed methods. Also we have to prove that our proposed techniques are more efficient for load balancing and energy consumption on cloud as compared to previous. IV. RESEARCH OBEJECTIVES The Objective of the proposed work is to balance the load and energy efficient using Dynamic load balancing model for cloud computing architecture. The Objectives of the research work are: 1. To optimize the performance of cloud architecture. 2. To implement proposed method ACCLB using java programming and simulate on cloud computing environment using CloudSim toolkit. IJTC www. ijtc.org 352

7 3. To compare the ACCLB with exiting load balancing techniques such as Vector Dot and Join- Idle queue for energy efficiency and load balancing. 4. To analyze the behavior of ACCLB using following parameters- Energy Consumption Throughput Response time Processing Time Fault Tolerance Total Execution Time 5. To Evaluate the performance and behavior of proposed ACCLB load balancing technique by comparing it with existing load balancing methods such as vector Dot and Join-Idle queue. V. RESEARCH METHODOLOGY Input: Required parameter for cloudlets and virtual machines are taken from user. Output: Improves energy consumption and load balancing at cloud with better response time, total execution time and throughput. Fig.2 Research Methodology IJTC www. ijtc.org 353

8 REFERENCES [1] Simranjeet Sidhu & Er. Manish Mittal, A New Era To Balance The Load On Cloud Using Vector Dot Load Balancing Method in the International Journal Of Technology And Computing (Ijtc), Volume 2, Issue 4 April [2] Ebin Deni Raj and Dhinesh Babu L.D, A Two Pass Scheduling Policy based Resource allocation for MapReduce in the International Conference on Information and Communication Technologies, ICICT [3] Wei Deng, Fangming Liu, and Hai Jin (2014) Harnessing Renewable Energy in Cloud Datacenters: Opportunities and Challenges, IEEE Network. [4] Ching-Hsien Hsu, Kenn D. Slagter, Shih-Chang Chen, Yeh-Ching Chung (2014) Optimizing energy consumption with task consolidation in clouds,science Direct. [5] Siva Theja Maguluri, R. Srikant, Lei Ying (2014), Heavy traffic optimal resource allocation algorithms for cloud computing clusters, In Elsevier, pp [6] Young Myoung Ko, Yongkyu Cho (2014), A distributed speed scaling and load balancing algorithm for energy efficient data centers, In Elsevier, pp [7] Huangke Chen, Xiaomin Zhu, Hui Guo, Jianghan Zhu, Xiao Qin, Jianhong Wu (2014), Towards Energy-Efficient Scheduling for Real-Time Tasks under Uncertain Cloud Computing Environment, In Elsevier,pp [8] Saiqin Long, Yuelong Zhao (2014), Wei Chen, A three-phase energy-saving strategy for cloud storage systems, In Elsevier, pp [9] Tom Guérout, Samir Medjiah, Georges Da Costa, Thierry Monteil (2014), Quality of service modeling for green scheduling in Clouds, In Elsevier. [10] Bernardetta Addis, Danilo Ardagna, Antonio Capone, Giuliana Carello (2014), Energy-aware joint management of networks and Cloud infrastructures, In Elsevier Computer Networks, pp [11] Patrick Raycroft, Ryan Jansen, Mateusz Jarus, Paul R. Brenner (2014), Performance bounded energy efficient virtual machine allocation in the global cloud, Elsevier,Sustainable Computing: Informatics and Systems, pp [12] S.K. Tesfatsion, E. Wadbro, J. Tordsson (2014), A combined frequency scaling and application elasticity approach forenergy-efficient cloud computing,elsevier, Sustainable Computing: Informatics and Systems, pp [12] Nader Mohamed, JameelaAl-Jaroodi, AbdullaEid (2013), A dual-direction technique for fast file downloads with dynamic load balancing in the Cloud, In Elsevier Journal of Network and Computer Applications, pp [13] Mohamed Abu Sharkh, Manar Jammal, Abdallah Shami, and Abdelkader Ouda (2013), Resource Allocation in a Network-Based Cloud Computing Environment: Design Challenges, IEEE Communications Magazine. [14] Gulshan Soni, Mala Kalra Comparative Study of Live Virtual Machine Migration Techniques in Cloud in the International Journal of Computer Applications ( ) Volume 84 No 14, December [15] Nidhi Jain Kansal and Inderveer Chana (2012), Existing Load Balancing Techniques in Cloud Computing: A Systematic Re-View, in Journal of Information Systems and Communication ISSN: , E-ISSN: , Volume 3, Issue 1, pp IJTC www. ijtc.org 354