resource allocation using multitier application in cloud computing

3 downloads 68 Views 482KB Size Report
Cloud computing is builds by the virtualization and distributed computing. On the bases ... The multitier application is actually for the infrastructure management.
International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963




Student M.Tech (CSE), 2Assistant Professor Lovely Professional University, Phagwara, Punjab, India

ABSTRACT Cloud computing is builds by the virtualization and distributed computing. On the bases of these factor cloud computing support cost- efficient usage of computing resources emphasizing an resource scalability and also on demand services. Now we are moving towards the advance communication and computational services that fulfill the all requirements of the users and also maintain the quality of services (QoS). There are necessary that the computing and networking resources need to be jointly treated and also optimized. For this purpose that is need of virtual resources, dynamically allocation of resources over the whole network. The dynamically allocation is the helpful and useful technique for handle the virtualized, Multitier application in the data center (cloud computing). V- Cache and service bus are useful for to manage overload at each tier in multitier application and also provide resources efficiently. The multitier application is actually for the infrastructure management. The network cloud mapping is actually the efficient mapping resources request on to a shared substrate. KEYWORDS: Quality of service, Infrastructure as a service, Platform as a service, Software as a service, Service level agreement, Service level agreement, Virtual network.



On demand resource provisioning management and quality of services (QoS) management is based on the virtual machines. In the cloud computing there are three types of services that are software as a service (SaaS), infrastructure as a service (IaaS), and platform as a services (PaaS). These all services have a very different business value proposition. Firstly traditional model came into existence related to the allocation of the resources. As this model have certain advantages and certain disadvantages also. In this model there was no on-demand allocation of resources and the workload factor was ignored which results in the slow processing. This model works on dedicated resources which means that only limited resources where allocated to it and due to these factors the processing rate degrades.

Figure 1: Traditional service computing framework


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963 Secondly there came a new model which replaced the previous model and tried to overcome the problems which were in the existing model. This model consists of two layers. The second layer added to this model is known as the virtual resource layer. The resources were distributed over these layers. The drawback of this model is also that it also lack on the working of workload. The factor of workload was also ignored which was an obstacle all though it is a 2-tier model.

Figure 2: capacity service computing framework

If consider the main goal of cloud computing then it can be said that the cloud computing creates large number of virtual resources, data centres and also servers those provide advantages to the user to access stored data and application according to their requirements. First thing the user demands is the reduction of cost. The IaaS provide the on- demand and immediate access of the computing resources with the cost saving for the users. So the capacity of the physical resources can be multiplexed among requested resources. A set of on- demand resource allocation algorithm is proposed, based on previous dynamic resource allocation mechanism with addition of SLA. This model consist service bus and cache tier. Cache tier is actually the machine learning based approach. The cache tier changes the intensity and scalability of the multitier application and also increases the throughput. Web tier and application tier are not directly joined. Service bus applied between the web tier and application tier. It consist the temporal decoupling, load balancing and load leveling. There are many challenges in on- demand resources dynamic provisioning for data centers like network bandwidth, partitioned among the VMs, disk, memory and also configuration for VMs. Cloud computing and networking create a deep relationship for the cloud. The network performance acts as a key for the cloud computing performance, so we can also say that there is a relationship between performance and resources provisioning of virtualized application. According to the research work is to be done on increase the performance of the cloud computing. Create a relationship between the dynamic provisioning models, virtualized multitier application and also resources mapping procedure. When join all these, then clusters of VMs are formed that are dedicated to virtualized multitier application and the dynamic provisioning models that determine how many VMs are allocated to virtualized multitier application to satisfy the end user request in a particular time period. According to this scenario the performance is high and the capacity is low, but our need is to optimize the both computing resources as well as the network resources. The computing resources act like server and network resources act as bandwidth. Also consider the functional and non-functional parameters, the functional parameters include the characteristics and properties of networking and computing (operating system, virtualization environment), the non-functional parameters include the criteria and constraints of the various resources like maximum disk space, maximum number of interfaces for each node and so on. The service provisioning in the cloud is based on the SLA (service level agreement) and it include the non-functional parameters. SLA requires the scheduling on requirement of CPU, network, storage, and bandwidth.


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963



Avinash Mehta, Mukesh Menaria, Sanket Dangi and Shrisha Rao, Energy Conservation in Cloud Infrastructures International Institute of Information Technology, Bangalore, IEEE (2011) [7]. This paper proposes the service request prediction model to achieving energy conservation in existing cloud infrastructure. The work of the service request prediction model is to determine the predefined period of time in which the server cluster will be under utilized. In this model they also define the load balancing mechanism. In this mechanism they accumulates all the requests, rather than distributing the load. This model also provides the less SLA violation with energy conservation. It reduces the overall cost and increases the lifetime of infrastructure. Jianfeng Yan Wen-Syan Li SAP Technology Lab, China Shanghai, Calibrating Resource Allocation for Parallel Processing of Analytic Tasks , 2009 IEEE International Conference on e-Business Engineering China [13], In this paper they described the challenge for the automated calibration of resource allocation for parallel processing and proposed an algorithm. This algorithm represented runtime statistic information and also calibrate the resource allocation accordingly. The experimental result of this algorithm describes that this algorithm is faster and more precision as compare to the other well know algorithms and also the pervious proposed algorithms. Jinhua Hu, Jianhua Gu et all, A Scheduling Strategy on Load Balancing of Virtual Machine Resources in Cloud Computing Environment, 3rd International Symposium on Parallel Architectures, Algorithms and Programming, 2010 IEEE [4], In this paper they described how balance the load on the VMs resources. For this work they purpose genetic algorithm. This algorithm increases load balancing factor and reduce the dynamic migration and high migration cost. This algorithm performs better even load is stable variant [4]. In this paper they also used the mapping between the VMs and physical machines for load balancing. Karthik Kumar, Jing Feng, Yamini Nimmagadda, and Yung-Hsiang Lu, Resource Allocation for Real-Time Tasks using Cloud Computing, School of Electrical and Computer Engineering, Purdue University, West Lafayette, 2011 IEEE [5], According to this paper they purposed the method to allocate the resources for real- time tasks. They use the infrastructure as a service model. There is a condition; the real time task has to be completed in the particular time period and also before the deadline. For this problem they purpose a scheme that is EDF- greedy scheme. According to this scheme they consider the temporal overlapping to allocate resources efficiently. Kazuki MOCHIZUKI† and Shin-ichi KURIBAYASHI, Evaluation of optimal resource allocation method for cloud computing environments with limited electric power capacity, 2011 International Conference on Network-Based Information Systems [8], In this paper they says that the limitation on the “electric power capacity” is major concept in each area, so they focus on that how they allocate the resources to the cloud computing with the limited electric power capacity. They say that: a. Network bandwidth and processing ability both are allocated simultaneously. b. They also purpose a method for optimally allocating the bandwidth and processing ability as well as the electric power capacity. c. They also purpose an algorithm for the electric power consumption. This algorithm reduces the electric power consumption by aggregating requests of multiple areas. Tino Schlegel, Ryszard Kowalczyk, Quoc Bao, Decentralized Co-Allocation of Interrelated Resources in Dynamic Environments, 2008 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology [11], In this paper they mentioned the decentralized co- allocation of interrelated resources in dynamic environment and it also includes the repeated jobs in real time. There is a resource broker agent that is autonomously allocating the resources for the execution of jobs but allocates the resources by the resource broker agent. It is based on the individual feedback and that feedback is received from the previous resource allocation decision. The result of this algorithm is very good and efficient for the open and dynamic environment with real application. There is a factor deadlock that may occur in between the agents, so for this factor they also purpose randomising techniques. They say that a limitation is set on the number of suitable resources providers in each broker. T.R. Gopalakrishnan Nair, Vaidehi M, Efficient resource arbitration and allocation strategies in cloud computing through, IEEE CCIS2011 [9], in this paper they purposed an algorithm that is rule based resource allocation (RBRA). This algorithm is based on the queuing model, means it is based on the


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963 priority management and also the FIFO approach that is first in first out approach. It can be said that, there is optimal resource allocation which is occurring if the rate of resource request from all subscribers is less than the rate with which the resource is allocated to subscribers. Vincent C. Emeakaroha, Ivona Brandic, Michael Maurer, Ivan Breskovic, SLA-Aware Application Deployment and Resource Allocation in Clouds, 2011 35th IEEE Annual Computer Software and Applications Conference Workshops [1], There is a one parameter known as SLA that is considered. In this paper they describe the multiple SLA parameters for deploying application in clouds. They define the heuristic design and implementation also. The heuristic design includes the load balancing mechanism. They also include the flexible on- demand resources usage in the heuristic. The aim of heuristic scheduling is to schedule applications on VMs with SLA terms and deployment. VMs on physical resources are totally based on resource availability. Hao Li, Jianhui Liu,Guo Tang, A Pricing Algorithm for Cloud Computing Resources, 2011 International Conference on Network Computing and Information Security [9]. This paper focuses on the scheduling and optimization of physical resources but they can’t provide the physical resources without the economic principles in the cloud applications. They purposed a cloud banking model. They consider the operating mechanism in banks, classification and quantification for the cloud resources, quality of services, and quality of use of cloud resources parameter. They also defined the pricing algorithm. The core of algorithm is CRP, CRP describe the following services a. It obtains the described tasks from the agent and participates in the competition. b. Calculate the total cost. c. It sends cost to the agents. d. It receives the user’s information from agency; implement the tasks from user and gets benefit. Chrysa Papagianni, Aris Leivadeas, Symeon Papavassiliou, Vasilis Maglaris, Cristina Cervello´ Pastor, and _ Alvaro Monje, On the Optimal Allocation of Virtual Resources in Cloud Computing Networks, IEEE transaction on computers, [3], Cloud computing is by building advances on virtualization and distributed computing to support cost-efficient usage of computing resources, emphasizing on resource scalability and on demand services. In this paper they are providing the unified resources allocation framework for networked clouds. They firstly formulated the optimal networked cloud mapping problem as a MIP (mixed integer programming problem). Efficiently mapping of resource requests onto a shared substrate interconnecting various islands of computing resources and adopt a heuristic methodology [3]. IaaS provides the on demand and immediate resources, actually the computing resources with the cost saving according to the user. Cloud provides two keys as Cloud computing and Networking. Functional parameter defined characteristic and properties of computing/ networking resources, for example operating system, supporting virtualization environment. Non- functional parameter specifies the criteria and the constraint, for example maximize the number of interfaces for each node, maximum disk space at the end.



3.1 Scope of study Resource allocation is one of the current areas in cloud computing, where techniques are applied to distribute scarce resources. Resources are allocated in cloud considering numerous parameters such as high throughput, maximum efficiency, SLA aware, quality of service, minimum energy power consumption etc. The aim of resource allocation system in cloud computing is to be sure about the applications requirements that are correctly attended by the provider’s infrastructure. 70% of Americans will be getting benefited from cloud and from its various applications by using email and connecting to social media through smart phones, watching movies over smart phones and uploading and accessing pictures from websites. Cloud computing is the major concept of our day- to- day life. Cloud computing is also the type of internet [10]. There is no doubt; presence of internet will boost its future. Cloud computing will becomes more important with the high- speed, broadband internet. The increasing presence of internet (cloud computing) is opening vistas in education and healthcare. Uses these services with little cost but for this, there are many techniques, algorithms are necessary to implement. There are three agents in the cloud computing, clients, provider, and developer. They consider the provider agents, how the resources are provided to the clients in a sufficient time periods.


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963 Today every village is connected with the internet. Wireless internet services are offered through the help of satellite, but the speed is too slow sometimes. Even airlines are offering satellite based wi- fi services with the help of cloud. Our work is toward optimizing the cloud. The optimizing the cloud means to optimize the functional and non-functional parameters, network, also the computing resources and networking resources. Networking resources means the bandwidth etc and the functional parameters related with the properties and characteristics like in cloud computing cost saving, cloud computing removes the requirements of a company to invest in storage hardware and severs. If include the concept of mapping of resources and dynamically provisional of resources with multitier application under SLA and the cost saving is increased. Existing resource allocation methods mainly focus on either central global optimization or local optimization within a server, but with some limitations on the scalability of cloud. Cache tier is applied before web tier and service bus is applied between the web tier and application tier. The cache tier changes the intensity and scalability of the multitier application. Web tier and application tier are not directly joined. Service bus applied between the web tier and application tier. It increases the performance security and flexibility. For the scope of good quality of services the following points are consider. 1. Network 2. Its infrastructure 3. Capacity 4. Dynamically provisioning 5. Configuration 6. Reconfiguration 7. Optimization 8. Scalability Necessary to improve the quality of services and management of workload on cloud computing. For this purpose map the multitier application with caching, service bus and resource requests on to a shared substrate interconnecting various islands of computing resources. In general way if consider the multitier architecture (3- tier architecture) like as

Figure 3: three tier architecture

Web tier: this tier directly access by the user such a desktop, UI, web pages etc, also called clients. Application tier: This tier encapsulate the business logic (such as business rules and data validation) domain concept, data access logic etc. also called middle layer. Database tier: The application tier use to store the application data such as data server, mainframe or legacy system etc.

3.2 Problem formulation Resource allocation mechanism should also consider the current status of each resource in the cloud environment. When apply any algorithm for better allocation of physical and/ or virtual resources the aim is to minimize the operational cost of the cloud environment. The first problem in the resource allocation is if the request is coming then how the resources are modelled. A networked cloud environment and request mapping model was designed which we also call as hardware representation. In this model the requests are coming from the user and going to the applications. Node mapping and link mapping is used in this architecture is to allocate the resources virtually over the cloud.


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963

Figure 4: Hardware representation of on-demand resource allocation problem

A new aspect on on-demand resource allocation problem was introduced which is also known as software representation. It is 2-tier model architecture. The drawbacks of previous model were overcome by this model. The virtual resource layer virtually allocate the resources such as to CPU, memory etc. but again the problem that existed in this model is of the workload. Though this model is also two-tier architecture but still it is not able to overcome the problem of workload.

Figure 5: on- demand resource allocation problem

Problem in resource allocation is divided into the five categories. a. Resource modelling and description. b. Resource offering and treatment. c. Resource discovering and monitoring. d. Resource selection. e. Work load handle in each tier. When the resource allocation system is develop then the first question arises that how to describe the resource present in the cloud. The development of suitable resource model and description is the first challenge that resource allocation service must address. An RAS also focus the challenges of representing the application requirements called resource offering and treatment. The mechanisms for resource discovery and monitoring are an essential part of the system. Provider phases the problem grouped in the conceptual phase, where resources must be modeled according to the variety of services the cloud will provide and the type of resources that it will offer. Operational phase when request for resource arrive, the RAS should initiate resource discovery to determine if there are required resources available in the cloud to attend the request. [12]


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963 Table 1: conceptual phase and operational phase Conception phase

Operational phase

Resource modelling

Resource discovering and modelling

Resource offering and treatments

Resource selection and optimization Overload manage

The problem is how the cloud IaaS handles the request, efficiently, efficient mapping of user requests for virtual resources onto a shared substrate interconnecting isolated island of computing resources with multitier application, SLA, caching and service bus. The problem is to solve the real- time problem of mapping virtual resources to substrate resources with limited assets. The virtual nodes and virtual links are known as virtual network embedding, but the main problem is, how the overload on each tier is distributed efficiently.

3.3 Objectives 1. Our work towards optimizing the cloud. Refer to this problem as SLA based optimally allocation of virtual resources for multitier application in cloud computing. 2. VNE algorithm suffers from scalability issues and hence request partitioning has been studied for mapping each part of the substrate network [3]. The resources always greater than the requests. 3. Networking performance related metrics can be further viewed as objectives that need to be optimized and/or constraints that need to be satisfied. For instance, one feasible way to reduce delay along a communication path is by minimizing transit. 4. Manage the workload on the each tier by applying some models and methods such as service bus, and caching model. 5. Workload and on- demand resources allocation must be familiar with each other so that the quality of application is not degraded, so both these must be tightly coupled with each other to prevent any inefficient usage.

3.4 Research methodology The resource elasticity is offered by the IaaS clouds for open opportunities for elastic application performance. But also poses challenges to application management. The management handles not only the capacity planning but also the proper partitioning of the resources into a number of virtual machines and also handles the workload on each tier. In the multitier architecture if adding a caching tier before the web tier then this caching tier boots application performance and reduce resource usages. Cache is actually a machine learning based approach. It identifies the incoming and dynamically resizes the cache space to accommodate these requests and used to optimally allocate the remaining capacity to the other tiers. The cache- tier change the intensity and traffic at static content and by caching tier the missing rate is lowest but the dynamic contents has high miss rate because the expiration of their time- to- live (TTL) values [10]. Combine the cache size and caching policy on application performance and combine resource usage of all tiers by which the performance of application is increased. They give the effective throughput; CPU, memory consumption etc. Apply the service bus between the web tier and application tier. The web tier and application tier are not connected directly to each other, it pushes units of work, until the application tier is ready to consume and process requests. By the indirectly messaging between them like

Figure 6: service bus between the wed and application tier

a. Temporal decoupling b. Load levelling c. Load balancing


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963 Temporal decoupling- It refers asynchronous messaging pattern, producers and consumers need not to store messages until the consuming party is ready to receive them. [14] Load levelling- In many application, system load varies over time whereas the processing time required for each unit of work is typically constant but by the intermediate message produce with queue means that provisioned to accommodate average load rather than peak load. It leads to the growth of queue and the contraction as well as the load changes. [14] Load balancing- as load increases, more work processes can be added to read from the queue. Further, more this pull- based load balancing allows for optimum utilization of work machines even if the work machines differ in term of processing power. [14]

Terms of processing power, as they view pull messages at their own maximum rate. So if join the caching tier with service bus, then the work load is managed across the all tiers.

Figure 8: The architecture of cache and service bus

Workload Analyzer define the request type, content size, the processing cost, response time, cache hit rate. Policy generator identifies the set of request that benefits most from caching, determine the minimum size of cache, redirected map provider, memory size of the caching tier is determined, and minimize the overall processing cost. Request redirector determine the request fall in cluster or not, if request fall in the cluster that is mapped to the cache server, the request redirector forwards it to caching tier otherwise the request is sent to the web tier. Resource manager allocates the remaining resources to all the tiers considering the overall performance of the multitier websites, the CPU allocation is managed.


Vol. 7, Issue 2, pp. 544-552

International Journal of Advances in Engineering & Technology, May, 2014. ©IJAET ISSN: 22311963


Future work

In the future, we will propose the algorithms that are control and increase the on- demand resource allocation among VMs. Determine the efficiency and potential of each tier, we can also extend the Vcache for heterogeneous application and control overload.

REFERENCES [1]. Emeakaroha, Vincent C., Ivona Brandic, Michael Maurer, and Ivan Breskovic. "SLA-Aware application deployment and resource allocation in clouds." In Computer Software and Applications Conference Workshops (COMPSACW), 2011 IEEE 35th Annual, pp. 298-303. IEEE, 2011. [2]. Goudarzi, Hadi, and Massoud Pedram. "Multi-dimensional sla-based resource allocation for multi-tier cloud computing systems." In Cloud Computing (CLOUD), 2011 IEEE International Conference on, pp. 324331. IEEE, 2011. [3]. Guo, Yanfei, Palden Lama, Jia Rao, and Xiaobo Zhou. "V-Cache: Towards Flexible Resource Provisioning for Multi-tier Applications in IaaS Clouds." [4]. Hu, Jinhua, Jianhua Gu, Guofei Sun, and Tianhai Zhao. "A scheduling strategy on load balancing of virtual machine resources in cloud computing environment." In Parallel Architectures, Algorithms and Programming (PAAP), 2010 Third International Symposium on, pp. 89-96. IEEE, 2010. [5]. Kumar, Karthik, Jing Feng, Yamini Nimmagadda, and Yung-Hsiang Lu. "Resource allocation for real-time tasks using cloud computing." In Computer Communications and Networks (ICCCN), 2011 Proceedings of 20th International Conference on, pp. 1-7. IEEE, 2011. [6]. Li, Hao, Jianhui Liu, and Guo Tang. "A pricing algorithm for cloud computing resources." In Network Computing and Information Security (NCIS), 2011 International Conference on, vol. 1, pp. 69-73. IEEE, 2011. [7]. Mehta, Avinash, Mukesh Menaria, Sanket Dangi, and Shrisha Rao. "Energy conservation in cloud infrastructures." In Systems Conference (SysCon), 2011 IEEE International, pp. 456-460. IEEE, 2011. [8]. Mochizuki, Kazuki, and Shin-ichi Kuribayashi. "Evaluation of optimal resource allocation method for cloud computing environments with limited electric power capacity." In Network-Based Information Systems (NBiS), 2011 14th International Conference on, pp. 1-5. IEEE, 2011. [9]. Nair, TR Gopalakrishnan, and M. Vaidehi. "Efficient resource arbitration and allocation strategies in cloud computing through virtualization." In Cloud Computing and Intelligence Systems (CCIS), 2011 IEEE International Conference on, pp. 397-401. IEEE, 2011. [10]. Papagianni, Chrysa, Aris Leivadeas, Symeon Papavassiliou, Vasilis Maglaris, and A. Monje. "On the optimal allocation of virtual resources in cloud computing networks." (2013): 1-1. [11]. Schlegel, Tino, Ryszard Kowalczyk, and Quoc Bao Vo. "Decentralized co-allocation of interrelated resources in dynamic environments." In Web Intelligence and Intelligent Agent Technology, 2008. WI-IAT'08. IEEE/WIC/ACM International Conference on, vol. 2, pp. 104-108. IEEE, 2008. [12]. Song, Ying, Yuzhong Sussn, and Weisong Shi. "A Two-Tiered On-Demand Resource Allocation Mechanism for VM-Based Data Centers." (2013): 1-1. [13]. Yan, Jianfeng, and Wen-Syan Li. "Calibrating Resource Allocation for Parallel Processing of Analytic Tasks." In e-Business Engineering, 2009. ICEBE'09. IEEE International Conference on, pp. 327-332. IEEE, 2009. Web References: [14]. [15]. and_Research_Challenges


Vol. 7, Issue 2, pp. 544-552

Suggest Documents