ENERGY-AWARE LOAD BALANCING AND APPLICATION SCALING FOR THE CLOUD ECOSYSTEM
Download as DOCX, PDF0 likes154 views
This document proposes a double resource renting scheme for cloud service providers to maximize profit while guaranteeing quality of service. It models the cloud system as an M/M/m+D queuing model. The scheme combines long-term and short-term server rentals to provide the necessary computing capacity over time. An optimization problem is formulated to determine the optimal server configuration that maximizes profit by balancing rental costs against increased revenue from meeting quality guarantees. Comparisons show the double renting scheme achieves higher profit than single renting while guaranteeing all requests are served on time.
![solutions and the actual optimal solutions. In addition, a series of calculations are conducted to
compare the profit obtained by the DQG renting scheme with the Single-Quality-Unguaranteed
(SQU) renting scheme. The results show that our scheme outperforms the SQU scheme in terms
of both of service quality and profit. In this paper, we only consider the profit maximization
problem in a homogeneous cloud environment, because the analysis of a heterogeneous
environment is much more complicated than that of a homogenous environment. However, we
will extend our study to a heterogeneous environment in the future.
REFERENCES
[1] K. Hwang, J. Dungaree, and G. C. Fox, Distributed and Cloud Computing. Elsevier/Morgan
Kaufmann, 2012.
[2] J. Cao, K. Hwang, K. Li, and A. Y. Zomaya, “Optimal multiserver configuration for profit
maximization in cloud computing,” IEEE Trans. Parallel Distrib. Syst., vol. 24, no. 6, pp. 1087–
1096, 2013.
[3] A. Fox, R. Griffith, A. Joseph, R. Katz, A. Konwinski, G. Lee, D. Patterson, A. Rabkin, and
I. Stoica, “Above the clouds: A berkeley view of cloud computing,” Dept. Electrical Eng. and
Comput. Sciences, vol. 28, 2009.
[4] R. Buyya, C. S. Yeo, S. Venugopal, J. Broberg, and I. Brandic, “Cloud computing and
emerging it platforms: Vision, hype, and reality for delivering computing as the 5th utility,”
Future Gener. Comp. Sy., vol. 25, no. 6, pp. 599– 616, 2009.
[5] P. Mell and T. Grance, “The NIST definition of cloud computing. National institute of
standards and technology,” Information Technology Laboratory, vol. 15, p. 2009, 2009.
[6] J. Chen, C. Wang, B. B. Zhou, L. Sun, Y. C. Lee, and A. Y. Zomaya, “Tradeoffs between
profit and customer satisfaction for service provisioning in the cloud,” in Proc.20th Int’l Symp.
High Performance Distributed Computing. ACM, 2011, pp. 229–238.
[7] J. Mei, K. Li, J. H, S. Yin, and E. H.-M. Sha, “Energy aware preemptive scheduling
algorithm for sporadic tasks
on dvs platform,” MICROPROCESS MICROSY., vol. 37, no. 1, pp. 99–112, 2013.](https://image.slidesharecdn.com/23-151112080838-lva1-app6892/85/ENERGY-AWARE-LOAD-BALANCING-AND-APPLICATION-SCALING-FOR-THE-CLOUD-ECOSYSTEM-6-320.jpg)
![[8] P. de Langen and B. Juurlink, “Leakage-aware multiprocessor scheduling,” J. Signal Process.
Sys., vol. 57, no. 1, pp. 73–88, 2009.
[9] G. P. Cachon and P. Feldman, “Dynamic versus static pricing in the presence of strategic
consumers,” Tech. Rep., 2010.
[10] Y. C. Lee, C. Wang, A. Y. Zomaya, and B. B. Zhou, “Profit driven scheduling for cloud
services with data access awareness,” J. Parallel Distr. Com., vol. 72, no. 4, pp. 591– 602, 2012.](https://image.slidesharecdn.com/23-151112080838-lva1-app6892/85/ENERGY-AWARE-LOAD-BALANCING-AND-APPLICATION-SCALING-FOR-THE-CLOUD-ECOSYSTEM-7-320.jpg)
ENERGY-AWARE LOAD BALANCING AND APPLICATION SCALING FOR THE CLOUD ECOSYSTEM
- 1. A PROFIT MAXIMIZATION SCHEME WITH GUARANTEED QUALITY OF SERVICE IN CLOUD COMPUTING Abstract—As an effective and efficient way to provide computing resources and services to customers on demand, cloud computing has become more and more popular. From cloud service providers’ perspective, profit is one of the most important considerations, and it is mainly determined by the configuration of a cloud service platform under given market demand. However, a single long-term renting scheme is usually adopted to configure a cloud platform, which cannot guarantee the service quality but leads to serious resource waste. In this paper, a double resource renting scheme is designed firstly in which short-term renting and long-term renting are combined aiming at the existing issues. This double renting scheme can effectively guarantee the quality of service of all requests and reduce the resource waste greatly. Secondly, a service system is considered as an M/M/m+D queuing model and the performance indicators that affect the profit of our double renting scheme are analyzed, e.g., the average charge, the ratio of requests that need temporary servers, and so forth. Thirdly, a profit maximization problem is formulated for the double renting scheme and the optimized configuration of a cloud platform is obtained by solving the profit maximization problem. Finally, a series of calculations are conducted to compare the profit of our proposed scheme with that of the single renting scheme. The results show that our scheme can not only guarantee the service quality of all requests, but also obtain more profit than the latter. EXISTING SYSTEM: In this section, we review recent works relevant to the profit of cloud service providers. Profit of service providers is related with many factors such as the price, the market demand, the system configuration, the customer satisfaction and so forth. Service providers naturally wish to set a higher price to get a higher profit margin; but doing so would decrease the customer satisfaction, which leads to a risk of discouraging demand in the future. Hence, selecting a reasonable pricing strategy is important for service providers. The pricing strategies are divided into two categories, i.e., static pricing and dynamic pricing. Static pricing means that the price of a service request is fixed and known in advance, and it does not change with the conditions. With dynamic pricing a
- 2. service provider delays the pricing decision until after the customer demand is revealed, so that the service provider can adjust prices accordingly. Static pricing is the dominant strategy which is widely used in real world and in research. Ghamkhari et al. Adopted a flat-rate pricing strategy and set a fixed price for all requests, but Odlyzko argued that the predominant flat-rate pricing encourages waste and is incompatible with service differentiation. Other kind of static pricing strategies are usage-based pricing. For example, the price of a service request is proportional to the service time and task execution requirement (measured by the number of instructions to be executed), respectively. Usage-based pricing reveals that one can use resources more efficiently. PROPOSED SYSTEM: In this paper, we propose a novel renting scheme for service providers, which not only can satisfy quality-of-service requirements, but also can obtain more profit. Our contributions in this paper can be summarized as follows. _ A novel double renting scheme is proposed for service providers. It combines long-term renting with short-term renting, which can not only satisfy quality-of-service requirements under the varying system workload, but also reduce the resource waste greatly. _ A multiserver system adopted in our paper is modeled as an M/M/m+D queuing model and the performance indicators are analyzed such as the average service charge, the ratio of requests that need short-term servers, and so forth. _ The optimal configuration problem of service providers for profit maximization is formulated and two kinds of optimal solutions, i.e., the ideal solutions and the actual solutions, are obtained respectively. _ A series of comparisons are given to verify the performance of our scheme. The results show that the proposed Double-Quality-Guaranteed (DQG) renting scheme can achieve more profit than the compared Single-Quality-Unguaranteed (SQU) renting scheme in the premise of guaranteeing the service quality completely. In this paper, to overcome the shortcomings mentioned above, a double renting scheme is designed to configure a cloud service platform, which can guarantee the service quality of all requests and reduce the resource waste greatly.
- 3. Moreover, a profit maximization problem is formulated and solved to get the optimal multiserver configuration which can product more profit than the optimal configuration. Module 1 A Cloud System Model The cloud structure consists of three typical parties, i.e., infrastructure providers, service providers and customers. This three-tier structure is used commonly in existing literatures. In the three-tier structure, an infrastructure provider the basic hardware and software facilities. A service provider rents resources from infrastructure providers and prepares a set of services in the form of virtual machine (VM). Infrastructure providers provide two kinds of resource renting schemes, e.g., long-term renting and short-term renting. In general, the rental price of long-term renting is much cheaper than that of short-term renting. A customer submits a service request to a service provider which delivers services on demand. The customer receives the desired result from the service provider with certain service-level agreement, and pays for the service based on the amount of the service and the service quality. Service providers pay infrastructure providers for renting their physical resources, and charge customers for processing their service requests, which generates cost and revenue, respectively. The profit is generated from the gap between the revenue and the cost. Module 2 A Multiserver Model In this paper, we consider the cloud service platform as a multiserver system with a service request queue. The schematic diagram of cloud computing. In an actual cloud computing platform such as Amazon EC2, IBM blue cloud, and private clouds, there are many work nodes managed by the cloud managers such as Eucalyptus, Open Nebula, and Nimbus. The clouds provide resources for jobs in the form of virtual machine (VM). In addition, the users submit their jobs to the cloud in which a job queuing system such as SGE, PBS, or Condor is used. All jobs are scheduled by the job scheduler and assigned to different VMs in a centralized way. Hence, we can consider it as a service request queue. For example, Condor is a specialized workload management system for compute intensive jobs and it provides a job queuing
- 4. mechanism, scheduling policy, priority scheme, resource monitoring, and resource management. Users submit their jobs to Condor, and Condor places them into a queue, chooses when and where to run they based upon a police. Hence, it is reasonable to abstract a cloud service platform as a multiserver model with a service request queue, and the model is widely adopted in existing literature. In the three-tier structure, a cloud service provider serves customers’ service requests by using a multiserver system which is rented from an infrastructure provider. Assume that the multiserver system consists of m long-term rented identical servers, and it can be scaled up by temporarily renting short-term servers from infrastructure providers. The servers in the system have identical execution speed s (Unit: billion instructions per second). In this paper, a multiserver system excluding the short-term servers is modeled as an M/M/m queuing system as follows (see Fig. 3). There is a Poisson stream of service requests with arrival rate λ, i.e., the interracial times are independent and identically distributed (i.i.d.) exponential random variables with mean 1/λ. A multiserver system maintains a queue with infinite capacity. When the incoming service requests cannot be processed immediately after they arrive, they are firstly placed in the queue until they can be handled by any available server. The first-come-first-served (FCFS) queuing discipline is adopted. The task execution requirements (measured by the number of instructions) are independent and identically distributed exponential random variables r with mean r (Unit: billion instructions). Therefore, the execution times of tasks on the multiserver system are also i.i.d. exponential random variables x = r/s with mean x = r/s (Unit second). The average service rate of each server is calculated as μ = 1/x = s/r, and the system utilization is defined as ρ = λ/mμ = λ/m _ r/s. Because the fixed computing capacity of the service system is limited, some requests would wait for a long time before they are served. According to the queuing theory, we have the following theorem about the waiting time in an M/M/m queuing system. Module 3 The Proposed Scheme In this section, we first propose the Double-Quality- Guaranteed (DQG) resource renting scheme which combines long-term renting with short-term renting. The main computing capacity is provided by the long-term rented servers due to their low price. The short-term rented servers
- 5. provide the extra capacity in peak period. The proposed DQG scheme adopts the traditional FCFS queuing discipline. For each service request entering the system, the system records its waiting time. The requests are assigned and executed on the long-term rented servers in the order of arrival times. Once the waiting time of a request reaches D, a temporary server is rented from infrastructure providers to process the request. We consider the novel service model as an M/M/m+D queuing model. The M/M/m+D model is a special M/M/m queuing model with impatient customers. In an M/M/m+D model, the requests are impatient and they have a maximal tolerable waiting time. If the waiting time exceeds the tolerable waiting time, they lose patience and leave the system. In our scheme, the impatient requests do not leave the system but are assigned to temporary rented servers. Since the requests with waiting time D are all assigned to temporary servers, it is apparent that all service requests can guarantee their deadline and are charged based on the workload according to the SLA. Hence, the revenue of the service provider increases. However, the cost increases as well due to the temporarily rented servers. Moreover, the amount of cost spent in renting temporary servers is determined by the computing capacity of the long-term rented multiserver system. Since the revenue has been maximized using our scheme, minimizing the cost is the key issue for profit maximization. Next, the tradeoff between the long-term rental cost and the short-term rental cost is considered, and an optimal problem is formulated in the following to get the optimal long-term configuration such that the profit is maximized. CONCLUSIONS In order to guarantee the quality of service requests and maximize the profit of service providers, this paper has proposed a novel Double-Quality-Guaranteed (DQG) renting scheme for service providers. This scheme combines short-term renting with long-term renting, which can reduce the resource waste greatly and adapt to the dynamical demand of computing capacity. An M/M/m+D queuing model is build for our multiserver system with varying system size. And then, an optimal configuration problem of profit maximization is formulated in which many factors are taken into considerations, such as the market demand, the workload of requests, the server-level agreement, the rental cost of servers, the cost of energy consumption, and so forth. The optimal solutions are solved for two different situations, which are the ideal optimal
- 6. solutions and the actual optimal solutions. In addition, a series of calculations are conducted to compare the profit obtained by the DQG renting scheme with the Single-Quality-Unguaranteed (SQU) renting scheme. The results show that our scheme outperforms the SQU scheme in terms of both of service quality and profit. In this paper, we only consider the profit maximization problem in a homogeneous cloud environment, because the analysis of a heterogeneous environment is much more complicated than that of a homogenous environment. However, we will extend our study to a heterogeneous environment in the future. REFERENCES [1] K. Hwang, J. Dungaree, and G. C. Fox, Distributed and Cloud Computing. Elsevier/Morgan Kaufmann, 2012. [2] J. Cao, K. Hwang, K. Li, and A. Y. Zomaya, “Optimal multiserver configuration for profit maximization in cloud computing,” IEEE Trans. Parallel Distrib. Syst., vol. 24, no. 6, pp. 1087– 1096, 2013. [3] A. Fox, R. Griffith, A. Joseph, R. Katz, A. Konwinski, G. Lee, D. Patterson, A. Rabkin, and I. Stoica, “Above the clouds: A berkeley view of cloud computing,” Dept. Electrical Eng. and Comput. Sciences, vol. 28, 2009. [4] R. Buyya, C. S. Yeo, S. Venugopal, J. Broberg, and I. Brandic, “Cloud computing and emerging it platforms: Vision, hype, and reality for delivering computing as the 5th utility,” Future Gener. Comp. Sy., vol. 25, no. 6, pp. 599– 616, 2009. [5] P. Mell and T. Grance, “The NIST definition of cloud computing. National institute of standards and technology,” Information Technology Laboratory, vol. 15, p. 2009, 2009. [6] J. Chen, C. Wang, B. B. Zhou, L. Sun, Y. C. Lee, and A. Y. Zomaya, “Tradeoffs between profit and customer satisfaction for service provisioning in the cloud,” in Proc.20th Int’l Symp. High Performance Distributed Computing. ACM, 2011, pp. 229–238. [7] J. Mei, K. Li, J. H, S. Yin, and E. H.-M. Sha, “Energy aware preemptive scheduling algorithm for sporadic tasks on dvs platform,” MICROPROCESS MICROSY., vol. 37, no. 1, pp. 99–112, 2013.
- 7. [8] P. de Langen and B. Juurlink, “Leakage-aware multiprocessor scheduling,” J. Signal Process. Sys., vol. 57, no. 1, pp. 73–88, 2009. [9] G. P. Cachon and P. Feldman, “Dynamic versus static pricing in the presence of strategic consumers,” Tech. Rep., 2010. [10] Y. C. Lee, C. Wang, A. Y. Zomaya, and B. B. Zhou, “Profit driven scheduling for cloud services with data access awareness,” J. Parallel Distr. Com., vol. 72, no. 4, pp. 591– 602, 2012.