Chapter 3 Virtualization Model for Cloud Computing Environment This chapter introduces the concept of virtualization in Cloud Computing Environment along with need of virtualization, components and characteristics of virtualization. As Cloud Computing Environment is a kind of distributed system that comprises of number of interconnected virtual computers which are dynamically provisioned in an unified manner to the cloud users on demand based on Service Level Agreement (SLA) we have proposed a virtualization model for cloud including cloud broker layer and used a mapping approach to show the mapping of VMs onto physical host machines depending on the availability of the distributed resources. This chapter provides the virtualization model which is the backbone for rest of the chapters to deal with security issues of auction and auditing schemes meant for Cloud Computing Environment. 3.1 Need for Virtualization Virtualization is one of the cost and energy saving technology which allows abstraction of physical hardware to provide virtual resources in the form of Virtual Machine. Through virtualization, resources of multiple physical machines are aggregated and assigned to applications dynamically on demand. Therefore, Virtualization is defined as a key technology of Cloud Computing Environment [18]. Using virtualization, multiple OSs and multiple applications can run on a single server at the same time, thus increasing hardware flexibility and utilization. 3.2 Major Components of Virtualization Environment The main aim of the virtualization is to create logical interface by abstracting the underlying infrastructure. Some of the components of virtualization environment are discussed below:
Guest- Guest represents the system component that interacts with the virtualization layer rather than with the host. Host- Host represents the original environment where the guests are supposed to be managed. Virtualization layer-virtualization layer is responsible for recreating the same or different environment where the guest will operate. It mainly deals with computation, storage and network virtualization. Virtualized resources are presented in this layer. 3.3 Characteristics of Virtualization With reference to Cloud Computing Environment, some important characteristics of virtualization are discussed below: 3.3.1 Consolidation Dedicated single system for single application concept is eliminated by Virtualization. Through virtualization multiple OSs and multiple applications can run on the same server. Both old and new versions of OS are capable of deploying on the same platform without additional investment on hardware. 3.3.2 Development Flexibility Application developer can run and test their applications in heterogeneous OS environments using same virtual machine. In virtualization environment, different applications are isolated from each other in their respective virtual partition. 3.3.3 Migration and Cloning To dynamically balance the workload, Virtual Machines are migrated from one site to another. As a result of which, users can access updated hardware and make recovery from hardware failure. Cloned virtual machines can be easily deployed on both local and remote sites. Page 37
3.3.4 Stability and Security In virtualized environment, host OSs are capable of hosting multiple guests OSs along with multiple applications. Each virtual machine is isolated from other virtual machines and not at all interfering into each other's work which helps in achieving stability and security. 3.3.5 Para Virtualization Para virtualization is an important aspect of virtualization. In virtualized environment, guest OS can run on host OS with or without modification. If any modification is made to the operating systems to be familiar with Virtual Machine Manager, then this process is called as Para virtualization. 3.4 Proposed Virtualization Model for Cloud Computing Environment Virtualization is the key concept Cloud Computing Environment to provide services to its users dynamically in the form of Virtual Machines. In this chapter, we have proposed a virtualization model for CCE [22] that reflects all components of cloud reference architecture including cloud broker layer. A cloud reference architecture basically consists of four layer: Resource as a Service (RaaS) [49], Infrastructure as a Service (IaaS), Platform as a Service (PaaS) and Software as a Service (SaaS). Different components of the proposed model are discussed below and shown in the Figure 3.1. 3.4.1 Distributed Resources of Physical Hosts At the lowest level of cloud stack, there exists a pool of resources that belong to physical hosts in a distributed manner. Resources are nothing but the collection of processors, CPU cores, memory, storage which are provisioned to the users in the form of virtual machines (VMs) using resource virtualization techniques. Page 38
3.4.2 Hypervisor Monitoring Environment (HME) Over the physical hosts, there exists Hypervisor Monitoring Environment layer which consists of Hypervisor, resource virtualization, VMs and Hypervisor administrator. a) Hypervisor Hypervisor can be defined as a piece of computer software or hardware that can create and run Virtual Machines. Hypervisor controls the access of VMs. Different types of hypervisor are listed below [3]: Native Hypervisor: These types of hypervisors can directly reside over the hardware platform to provide better performance. Embedded or Bare Metal Hypervisor: These types of hypervisors are integrated with processors on a separate chip to achieve high performance and to reduce time complexity. Host Hypervisor: These types of hypervisors are represented as a separate layer of software over both operating system and hardware to improve performance. b) Virtualization layer Virtualization layer deals with network, storage and computation virtualization. Through virtualization, resources of multiple physical machines are aggregated and assigned to applications dynamically on demand in the form of Virtual Machines (VM). Virtual Machines are the software implementation of computing resources where an operating system or a program can be installed or run [50]. Virtual Machine is created within host computing environment using virtualization technique. Multiple VMs exist within a single host at the same time without interfering each other. The basic responsibility of VMs are to achieve improved scalability and performance for Page 39
the running application in Cloud. Virtualization reduces the cost by reducing the need for the physical hardware system. Different types of resource virtualization are listed below. Network virtualization In network virtualization, virtual networks are implemented within the physical server for switching between all the virtual servers instead of multipath or multiplexed network channels by directly trunking them to WAN transport and multiple HBAs and NICs may be needed for each application provided with a high speed single Ethernet connection [19]. Storage virtualization Fibre Channel (FC) and FC-based Storage Area Networks (FC-SAN) are the key element of storage virtualization which facilitates storage connectivity (very high speed) and storage solutions like point-to-point replication and server-less back-up [19]. It also helps to optimize the performance of the servers on which multiple applications are running and this technique has enabled numerous advancements in the data center. Computation virtualization Virtualization of computing resources like server and operating system has transformed the server centric traditional computing into network and Internet centric computing. Virtualization of operating system and server create the virtual (logical) servers which are free from constraints of the underlying physical location and physical infrastructure and facilitate the move of the workloads from source VM instance to target VM instance in real-time which is called Live Migration [19]. Page 40
Figure 3.1: Virtualization model for Cloud Computing Environment [22] Page 41
c) Hypervisor Administrator In CCE, the major responsibility of Hypervisor Administrator to control the access of VMs. Hypervisor Administrator monitors the distribution and circulation of computing resources for virtualization and the mapping of virtual machines to host machines so that service can be provided to the cloud user on demand. Hypervisor Administrator can manage resources located within the server and resources located in another server connected to the network [19]. 3.4.3 Platform Service Platform service can create FCAPS oriented (Fault, Configuration, Account, Performance and Security) virtual servers [19] which can host guest operating systems to execute multiple user- applications. Different components of this layer are listed below: a) Integration Service Platform Integration Service Platform is responsible for integration of cloud resources for different guest operating systems and applications. b) Platform and OS usage and services In this block, virtual servers load and execute the user's application by hosting appropriate operating systems. Application developers do not necessarily need to know the underlying infrastructure to create and deploy applications in a cloud platform. c) Performance Measurements and Update Configuration In this block, service developers are allowed to prepare self-configured and selfmanaged business workflows. Virtual servers are responsible for managing performance measurements, performance optimization, fault management, reliability and accountability. Page 42
3.4.4 Software Service Through software service layer, clients are provided with ready to use software service without taking any overhead for buying and maintaining the software. Different components of software service to facilitate the software creation, delivery and usage [19] are listed below: a) Services Creation Platform Services creation platform provides the tools to create different applications which can be distributed over the virtual servers. b) Services Delivery Platform Services delivery platform can be defined as a workflow engine which is capable of executing applications and managing the orchestration of multiple distributable workflow elements. c) Legacy Integration Services Legacy integration services are responsible for the inheritance or legacy of previous application. 3.4.5 Broker Service Cloud brokers act as agents between the virtual environment and business service layer. Optimal arrangement of services provided by different CSPs which is configured and managed by the service manager. There exist network broker, storage broker and computation broker to perform the respective brokering service. 3.4.6 Business Service Business controls and conditions like Business service management, Identity creation and validation, Authentication and Access Control Management, Service level Page 43
agreement, and Renewal service management are provided by the Business service layer. a) Business Service Management Service orchestration, service conditions and process management are managed by the Business service management block. Service manager is capable of deriving a collection and configuration of desired resources along with the constraints according to licensing, cost, confidentiality etc.. Service manager is responsible for monitoring the deployment of services according to their capacity (number of VM instances). Process management is responsible for scheduling and managing the processes. b) Identity Creation and Validation User identity is required for deployment of user-application in cloud. This block is responsible for creation and validation of user identity. c) Authentication and Access Control Management Authentication establishes the identity which is required for effective governance and management of different process involves in service provisioning. Access control management service determines who can access what in response to a request from a cloud user to get a required service. Authentication is a prerequisite for access control management service. d) Service Level Agreement (SLA) A Service-Level-Agreement (SLA) is a formal way to specify both functional and non-functional conditions under which different services are provided. SLA acts as a contract negotiated between Cloud Service Provider and cloud user. Both Cloud Service Provider and cloud user can refer SLAs to check whether the required service is provided as per the SLA terms or not. In case of any violation, SLAs allow for penalties to be derived directly. Page 44
e) Renewal Service Management Renewal service management is responsible for renewal of rules, polices, access controls, SLA update and validation. 3.5 Workflow of the Proposed Virtualization Model The workflow of the proposed virtualization model reflects the communication among cloud user, cloud broker and Cloud Service Provider as shown in Figure 3.2. Figure 3.2 : Work flow of the proposed virtualization model Page 45
3.6 A Mapping Approach for Virtual Machines A mapping approach is used to show the mapping of VMs onto physical host machines depending on the availability of the distributed resources [22][23]. In this thesis, we have represented the system as S, set of Virtual Machines as VS, set of physical host machines as HS, and the pool of physical resources as P. According to the user-requirements, VM instances are created by the hypervisor administrator. VS is defined as: VS=V 1 +V 2 +...+Vm= V i V i = {vc,vm,vr} where, vc= Number of CPU cores vm= Main memory vr = Storage capacity m= Number of Virtual Machines HS is defined as HS= H 1 + H 2 +...+ H n = H i H i = {hc, hm, hr} where, hc= Number of CPU cores hm= Main memory hr = Storage capacity n= Number of physical host machines HS was divided into two subsets HS a and HS b where a + b= n and HS a = Set of physical host machines with available resources to host VMs HS b = Set of physical host machines that do not have enough resources to host VMs The function f:v i HS a was used to map VM instances to the set of physical host machines having enough resources to host VMs. The mapping could be of one to one or many to one type. One VM instance can be mapped onto one physical host in one to one mapping represented by the function f:v i H i, where as many VM instances Page 46
can be mapped onto one physical host in many to one mapping represented by the function f: V i H i. 3.7 Performance Analysis Considering this virtualization model for cloud, we have selected different VM allocation and single VM selection policy. The simulation tests were conducted on a 32-bit Intel Core i5 machine having 2.60 GHz and 3 GB RAM running Windows 7 Professional and JDK 1.6 using CloudSim version 3.0[47]. During the simulation process, we have created a heterogeneous power aware datacenter that applied VM allocation and VM selection policies. We have considered two types of host: HP Proliant ML110 G4 Xeon 3040 machine having 1.86 GHz processor (1860 MIPS), Dual core and HP Proliant ML110 G5 Xeon 3075 machine having 2.66 GHz processor. (2660 MIPS), dual core. Both host machines are modelled as 4 GB of RAM, 1TB of Storage. In this simulation, we have used four types of VMs; each of VM having 2500, 2000, 1000 and 500MIPS and 870, 1740, 1740, and 613 MB of RAM respectively. Also each VM had a single core and 2.5GB of VM size. Datacenter used for this simulation had the characteristics like architecture of x86, Linux Operating System and Xen VMM. We have considered three types of VM Allocation policy for simulation point of view: Inter Quartile Range (Iqr), Median Absolute Deviation (Mad), and Static Threshold (Thr) [47][48] and Minimum Migration Time (Mmt) as VM Selection policy [47][48]. We have compared each VM Allocation policy with Minimum Migration Time while changing the numbers of VMs. Page 47
Figure 3.3: Experiment Results-Total Energy Consumption by the system Varying the numbers of VMs from 10 to 100 and we have calculated the energy consumption (in KWh) while considering three cases like IqrMmt, MadMmt, and ThrMmt. In the third case, where we have used Static Threshold as VM allocation policy and Minimum Migration Time as VM selection policy, the energy consumption is less than other two cases as shown in the Figure 3.3 [51]. Figure 3.4: Experiment Results-Total Execution Time by the System Page 48
In Figure 3.4 [51], we have presented the result for calculating total execution time. Here, we have considered the same three situations as stated above and here also we can say that if we use Static Threshold as VM allocation policy and Minimum Migration Time as VM selection policy, the total execution time is much more less than the other two situations. Summary We summarize our work as follows: This chapter focuses on different aspects of virtualization techniques and proposes a virtualization model for Cloud Computing Environment. This chapter also represents a mapping approach to show the mapping of VMs onto physical host machines depending on the availability of the distributed resources. The proposed virtualization model works as a backbone to understand and deal with the security issues of auction and auditing schemes in rest chapters. The next chapter discusses the issues of existing auction schemes for VM instances allocation and presents an auction based VM instance allocation scheme that achieves confidentiality of bid value. Page 49