Msrtia: A proposal to reduce the response time for load balancing on cloud computing
Abstract: Cloud computing is a model that provides
everything related to information technology in the form
of services through the Internet. The primary benefit of it
is to save the original system investment cost, optimize
the data processing, calculating and storing data.
Nowadays, cloud computing faces many challenges in
ensuring the quality of service throughout. In which the
problem of overloading physical servers or virtual servers
of data centres is concerned specially. So as to qualify the
above requests, setup an effective load balancing method
and using resources with the most optimization is the
target which cloud computing wants to gain. In this paper,
we propose Max-Min Scheduling Response Time
Improved Algorithm (MSRTIA) basing on Max Min
Scheduling algorithm. Our algorithm calculates the
Cloudlet aggregation value of requests then pairs the
request with the largest value found with the fastest
executing virtual machine (VM). In which, Cloudlet
aggregation value is the association of length, output size
and file size parameters. The simulation result proves that
MSRTIA has less response time in comparison with MaxMin Scheduling and Round-Robin algorithms.
Trang 1
Trang 2
Trang 3
Trang 4
Trang 5
Trang 6
Trang 7
Trang 8
Tóm tắt nội dung tài liệu: Msrtia: A proposal to reduce the response time for load balancing on cloud computing
E RESPONSE TIME FOR LOAD BALANCING ON CLOUD COMPUTING Figure 8. MSRTIA schema Description of the MSRTIA algorithm: The idea is to calculate the Cloudlet aggregation value of requests, chooses the request with largest value, then pairs with the fastest executing VM following the formula: public double getAverageSize(Cloudlet cloudlet) { double resAverage; resAverage = cloudlet.getCloudletLength() * 0.5 + cloudlet.getCloudletOutputSize() * 0.2 + cloudlet.getCloudletFileSize() * 0.3; return resAverage; } The MSRTIA algorithm will repeat until the Makespan tables are empty. At that time the requests will be processed faster, shorten the finishing time, increased load balancing capability for cloud computing. The response time is explained in details as below. Basis of assessment The efficiency of load balancing can be based on many factors, but the most important factors are loading and performance. Loading is the CPU queue index and utilization [15]. Performance is the average response time to user requests. Load balancing algorithm is based on input parameters such as: configuring virtual machines, configuring Cloudlet tasks, arrival time, and time to complete tasks, then estimating the expected completion time of each task. Response time is the processing time plus the cost of transmitting request, queuing through the network nodes. Expected response time is calculated according to the following formula [16]: Expected Response Time = F – A + Tdelay F: time to complete the task, A: arrival time of the task, Tdelay: transfer time of the task Because the algorithm that performs load balancing is that of DatacenterBroker, the level of the algorithm only affects the processing time in a local environment of a data centre. Hence the communication delay parameter can be omitted, so Tdelay = 0. Determining the expected time to complete task [16]: If the scheduling policy is Space shared – Space shared or Time shared - Space shared, it is determined by the following formula: eft(p) = est(p) + (3.1) capacity = (3.2) If the scheduling policy is Space shared – Time shared or Timeshared-Timeshared, it is determined by the formula eft(p) = ct + (3.3) capacity = (3.4) eft(p) is the expected completion time of Cloudlet p; est is the arrival time of Cloudlet p; rl is the total number of instructions that Cloudlet p needs to execute on a processor; capacity is the average processing power (in MIPS) of a core for Cloudlet p; ct is the current simulation time; cores (p) is the number of cores required by Cloudlet; np is the number of actual cores that the host is considering; cap is the processing power of the core. The capacity parameter specifies the actual capacity for task processing on each VM. Apparently capacity depends on the scheduling of computing resources on the virtualized system. The total processing power on a physical host is unchanged and depends on the number of physical cores and processing power of each core. However, when this processing resource is shared for multiple tasks simultaneously, each task requires a certain number of cores and if the total number of cores is greater than the number of physical cores, the concept of virtual core appears, each virtual core will have lower processing power than the physical core. In other words, the capacity of a virtual core for a task can only be equal to or smaller than the physical core and how much depends on the resource sharing policy. Capacity is the processing power of a virtual core [15] [16]. Tran Cong Hung, Nguyen Ngoc Thang, Kieu Trong Duc From this analysis and based on the resource sharing policy to develop formulas for capacity. Resource sharing policy is specified through scheduling mechanism in cloud computing. We have two levels of scheduling: scheduling virtual machines to share physical host machine resources and scheduling tasks to share virtual machine resources. There are two scheduling mechanisms: Time shared and Space shared. Within the scope of this paper, we will perform algorithms and simulations based on the Time shared – Time shared policy, respectively, to virtual machines and tasks. Therefore, the calculation base for the proposed algorithm will be based on the formulas (3.3) and (3.4). IV. SIMULATION & EVALUATION Cloud environment emulator uses CloudSim 3.0 library and programming in JAVA language, includes 1 to 4 VMs. It will create a random request environment for services on the cloud containing virtual cloud services, CloudSim provisioning and user provisioning services for testing [17]. Table 1. Data center configuration parameters Table 2. VM configurations parameters when initialized The requests (WebRequest) are represented by Cloudlet in CloudSim and the size of Cloudlets is randomly generated using the JAVA random function. Table 3. Requests configuration parameters The function to create randomly 1000 requests in Table 2: public DataInput() { this.v_length = ThreadLocalRandom.current().nextInt(1700, 3000); this.v_fileSize = ThreadLocalRandom.current().nextInt(5000,45000); this.v_outputSize= ThreadLocalRandom.current().nextInt(450,750); } Result and Evaluation Experiments apply Timeshared – Timeshared scheduling policy for VM – task and calculate response time according to formula (3.3) and (3.4) as described in Base of assessment. The simulation will make out 1000 requests with 4 times, each will have 4 VMs and the number of requests is 100, 200, 500 and 1000 respectively. From the Figure 9-13, we can see that the response time of MSRTIA is less than Max-Min and Round-Robin for all scenarios with the number of requests from 10 to 1000. The more requests are tested, the better response time of MSRTIA demonstrates in comparison with Max-Min and Round-Robin. In other words, MSRTIA is effective especially for large amount of requests. Through 4 experiments, it shows that MSRTIA has the response time for VMs better and load balancing more efficiently than Max-Min and Round-Robin scheduling techniques. Specifically, the response time of MSRTIA is faster 9.63% than Max-Min and 15.32% than Round- Robin algorithms. Compared to the previous methods, the proposed algorithm doesn’t need to perform the calculation for completion time of requests again. From which, MSRTIA will reduce unnecessary processing time and costs as well as minimize load unbalancing in the cloud system. Figure 9. Experimental result on 4 virtual machines with 100 requests MSRTIA: A PROPOSAL TO REDUCE THE RESPONSE TIME FOR LOAD BALANCING ON CLOUD COMPUTING Figure 10. Experimental result on 4 virtual machines with 200 requests Figure 11. Experimental result on 4 virtual machines with 500 requests Figure 12. Experimental result on 4 virtual machines with 1000 requests Figure 13. Experimental result after 4 times counted on average V. CONCLUSION MSRTIA calculates the Cloudlet aggregation value of requests and searches for the request owning the maximum value then assign to VM having the minimum completion time. It is very clear to see the results that MSRTIA has ameliorated the response time for load balancing, optimized the performance compared to Max- Min and Round-Robin algorithms with the better rate 9.63% and 15.32% respectively. For future research, the improvements may include the following: simulate the algorithm with more configuration cases in terms of data centres, VMs, different scheduling policies (Time shared – Space shared or vice versa), combined with other machine learning methods. REFERENCES [1] Agraj Sharma, Sateesh K. Peddoju, (2014), “Response Time Based Load Balancing in Cloud Computing”, International Conference on Control, Instrumentation, Communication and Computational Technologies (ICCICCT). [2] J. Zhao, K. Yang, X. Wei, Y. Ding, L. Hu and G. Xu, (2016) “A Heuristic Clustering-Based Task Deployment Approach for Load Balancing Using Bayes Theorem in Cloud Environment”, in IEEE Transactions on Parallel and Distributed Systems, vol. 27, no. 2, pp. 305-316. [3] J. Zhang, Q. Liu and J. Chen, (2016) “An Advanced Load Balancing Strategy for Cloud Environment”, International Conference on Parallel and Distributed Computing, Applications and Technologies (PDCAT), Guangzhou, pp. 240-243. [4] L. Pallavi, V. Pradeep Kumar, (2014), "Mobile Cloud Computing: Service Models", International Conference on Computer & Communication Technologies, INDIA. [5] Mohammad UbaidullahBokhari, Qahtan Makki Shallal, YahyaKordTamandani, (2016), "Mobile Cloud Computing Service Models: A Comparative Study", IEEE Network, Institute of Electrical and Electronics Engineers network. [6] Mohammad Riyaz Belgaum, Safeeullah Soomro, Zainab Alansari, Muhammad Alam, Shahrulniza Musa, Mazliham Mohd Su'ud, (2017), “Load Balancing with preemptive and non-preemptive task scheduling in Cloud Computing”, International Conference on Engineering Technologies and Social Sciences (ICETSS). [7] Divya Chaudhary, Rajender Singh Chhillar, (2013) “A New Load Balancing Technique for Virtual Machine Cloud Computing Environment”, International Journal of Computer Applications (0975 – 8887), Volume 69– No.23. Tran Cong Hung, Nguyen Ngoc Thang, Kieu Trong Duc [8] Soheil Anousha, Mahmoud Ahmadi, (2013), “An Improved Min-Min Task Scheduling Algorithm in Grid Computing”, Conference Paper Analysis and Tools for Integrated Circuits and Systems pp.103-113. [9] Poonam Kumari1, Mohit Saxena, (2016) “A Round- Robin based Load balancing approach for Scalable demands and maximized Resource availability”, International Journal of Engineering and Computer Science, ISSN: 2319-7242. Volume 5, Page No. 17375-17380. [10] Mustafa ElGili Mustafa, (2017) “Load Balancing Algorithms Round-Robin (RR), Least connection, And Least Loaded Efficiency”, GESJ: Computer Science and Telecommunications, No.1(51). [11] Shivangi Mayur, Nidhi Chaudhary, (2019) “Enhanced Weighted Round Robin Load Balancing Algorithm in Cloud Computing”, International Journal of Innovative Technology and Exploring Engineering (IJITEE), ISSN: 2278-3075, Volume-8, Issue- 9S2. [12] Nguyen Xuan Phi, Cao Trung Tin, Luu Nguyen Ky Thu and Tran Cong Hung, (2018), “Proposed Load Balancing Algorithm To Reduce Response Time And Processing Time On Cloud Computing”, International Journal of Computer Networks & Communications (IJCNC) Vol.10, No.3. [13] Mao Y., Chen X., Li X, (2014) “Max–Min Task Scheduling Algorithm for Load Balance in Cloud Computing”, In: Patnaik S., Li X. (eds) Proceedings of International Conference on Computer Science and Information Technology. Advances in Intelligent Systems and Computing, vol 255. Springer, New Delhi. [14] Bhavisha Kanani, Bhumi Maniyar, (2015) “Review on Max-Min Task Scheduling Algorithm for Cloud Computing”, JETIR, (ISSN-2349-5162), Volume 2, Issue 3. [15] Nguyen Xuan Phi and Tran Cong Hung, (2017) “Load Balancing Algorithm to improve response time on cloud computing”, International Journal on Cloud Computing: Services and Architecture (IJCCSA). Vol.7, No.6. [16] Rodrigo N. Calheiros, Rajiv Ranjan, Anton Beloglazov, C´ esar A. F. De Rose and Rajkumar Buyya, (2010) “CloudSim: a toolkit for modeling and simulation of cloud computing environments and evaluation of resource provisioning algorithms”, Software: Practice and Experience (SPE), Volume 41 Number 1, pp.23-50. [17] Tran Cong Hung, Phan Thanh Hy, Le Ngoc Hieu, Nguyen Xuan Phi, (2019) “Improved Max-Min Scheduling Algorithm for Load Balancing on Cloud Computing”, ICMLSC Proceedings of the 3rd International Conference on Machine Learning and Soft Computing, pp.60-64. MSRTIA: MỘT ĐỀ XUẤT ĐỂ GIẢM THỜI GIAN ĐÁP ỨNG CHO CÂN BẰNG TẢI TRÊN ĐIỆN TOÁN ĐÁM MÂY Tóm tắt: Điện toán đám mây là mô hình cung cấp mọi thứ liên quan đến công nghệ thông tin dưới dạng dịch vụ thông qua Internet. Lợi ích chính của nó là tiết kiệm chi phí đầu tư hệ thống ban đầu, tối ưu hóa việc xử lý dữ liệu, tính toán và lưu trữ dữ liệu. Ngày nay, điện toán đám mây phải đối mặt với nhiều thách thức trong việc đảm bảo chất lượng dịch vụ xuyên suốt. Trong đó vấn đề quá tải máy chủ vật lý hoặc máy chủ ảo của các trung tâm dữ liệu được đặc biệt quan tâm. Vì vậy, để đủ điều kiện cho các yêu cầu trên, thiết lập một phương pháp cân bằng tải hiệu quả và sử dụng các tài nguyên tối ưu hóa nhất là mục tiêu mà điện toán đám mây muốn đạt được. Trong bài báo này, chúng tôi đề xuất thuật toán MSRTIA dựa trên thuật toán lập lịch Max Min. Thuật toán của chúng tôi tính toán giá trị tổng hợp Cloudlet của các yêu cầu sau đó ghép yêu cầu với giá trị lớn nhất được tìm thấy với máy ảo thực thi nhanh nhất. Trong đó, giá trị tổng hợp của Cloudlet là sự kết hợp của các tham số độ dài, kích thước đầu ra và kích thước tệp. Kết quả mô phỏng chứng minh rằng MSRTIA có thời gian phản hồi nhanh hơn so với các thuật toán Max Min và Round-Robin. Từ khóa: Giá trị tổng hợp Cloudlet, Max-Min, Round- Robin AUTHORS Tran Cong Hung was born in Vietnam in 1961. He received the B.E in electronic and Telecommunication engineering with first class honors from HOCHIMINH University of technology in Vietnam, 1987. He received the B.E in informatics and computer engineering from HOCHIMINH University of technology in Vietnam, 1995. He received the Master of Engineering degree in telecommunications engineering course from postgraduate department Hanoi University of technology in Vietnam, 1998. He received PhD. at Hanoi University of technology in Vietnam, 2004. His main research areas are B – ISDN performance parameters and measuring methods, QoS in high speed networks, MPLS. He is, currently, Associate Professor PhD. of Faculty of Information Technology II, Posts and Telecoms Institute of Technology in HOCHIMINH, Vietnam. Nguyen Ngoc Thang was born in Vietnam in 1993. He received B.E in Computer Science with first class honors from Industrial University of Ho Chi Minh City, MSRTIA: A PROPOSAL TO REDUCE THE RESPONSE TIME FOR LOAD BALANCING ON CLOUD COMPUTING Vietnam, 2015. He received the Master of Computer Science degree with first class honors from Saigon University, Vietnam in 2018. Kieu Trong Duc was born in Vietnam in 1989. He received his undergraduate degree in 2011, major in Information Technology from Saigon University, Viet Nam. Currently, he is a Master candidate in Computer Science of Saigon University, Vietnam. He is working for the Vietnam Mobile Telecom Services One Member Limited Liability Company.
File đính kèm:
- msrtia_a_proposal_to_reduce_the_response_time_for_load_balan.pdf