Modeling and Simulation of Scalable Cloud Computing Environments and
the CloudSim Toolkit: Challenges and Opportunities
Abstract
Cloud computing aims to power the next generation datacenters and enables application service providers to leasedata center capabilities for deploying applicationsdepending on user QoS (Quality of Service) requirements.Cloud applications have different composition,configuration, and deployment requirements. Quantifyingthe performance of resource allocation policies andapplication scheduling algorithms at finer details in Cloudcomputing environments for different application andservice models under varying load, energy performance(power consumption, heat dissipation), and system size is achallenging problem to tackle. To simplify this process, inthis paper we propose CloudSim: an extensible simulationtoolkit that enables modelling and simulation of Cloudcomputing environments. The CloudSim toolkit supportsmodelling and creation of one or more virtual machines(VMs) on a simulated node of a Data Center, jobs, andtheir mapping to suitable VMs. It also allows simulation ofmultiple Data Centers to enable a study on federation andassociated policies for migration of VMs for reliability andautomatic scaling of applications.
1. Introduction
Cloud computing delivers infrastructure, platform, andsoftware as services, which are made available assubscription-based services in a pay-as-you-go model toconsumers. These services in industry are respectivelyreferred to as Infrastructure as a Service (IaaS), Platform asa Service (PaaS), and Software as a Service (SaaS). Theimportance of these services is highlighted in a recentreport from Berkeley as: “Cloud computing, the long-helddream of computing as a utility, has the potential totransform a large part of the IT industry, making softwareeven more attractive as a service” [11].Clouds [10] aim to power the next generation datacenters by exposing them as a network of virtual services(hardware, database, user-interface, application logic) sothat users are able to access and deploy applications fromanywhere in the world on demand at competitive costsdepending on users QoS (Quality of Service)requirements [1]. Developers with innovative ideas for newInternet services are no longer required to make largecapital outlays in the hardware and software infrastructuresto deploy their services or human expense to operate it[11]. It offers significant benefit to IT companies byfreeing them from the low level task of setting up basichardware and software infrastructures and thus enablingmore focus on innovation and creation of business values.Some of the traditional and emerging Cloud-basedapplications include social networking, web hosting,content delivery, and real time instrumented dataprocessing. Each of these application types has differentcomposition, configuration, and deployment requirements.Quantifying the performance of scheduling and allocationpolicies in a real Cloud environment for differentapplication and service models under different conditionsis extremely challenging because: (i) Clouds exhibitvarying demand, supply patterns, and system size; and (ii)users have heterogenous and competing QoS requirements.The use of real infrastructures such as Amazon EC2, limitsthe experiments to the scale of the infrastructure, andmakes the reproduction of results an extremely difficultundertaking. The main reason for this being the conditionsprevailing in the Internet-based environments are beyondthe control of developers of resource allocation andapplication scheduling algorithms.An alternative is the utilization of simulation tools thatopen the possibility of evaluating the hypothesis prior tosoftware development in an environment where one canreproduce tests. Specifically in the case of Cloudcomputing, where access to the infrastructure incurspayments in real currency, simulation-based approachesoffer significant benefits to Cloud customers by allowing2them to: (i) test their services in repeatable and controllableenvironment free of cost; and (ii) tune the performancebottlenecks before deploying on real Clouds. At theprovider side, simulation environments allow evaluation ofdifferent kinds of resource leasing scenarios under varyingload and pricing distributions. Such studies could aidproviders in optimizing the resource access cost with focuson improving profits. In the absence of such simulationplatforms, Cloud customers and providers have to relyeither on theoretical and imprecise evaluations, or on tryand-error approaches that lead to inefficient serviceperformance and revenue generation.Considering that none of the current distributed systemsimulators [4][7][9] offer the environment that can bedirectly used by the Cloud computing community, wepropose CloudSim: a new, generalized, and extensiblesimulation framework that enables seamless modeling,simulation, and experimentation of emerging Cloudcomputing infrastructures and application services. Byusing CloudSim, researchers and industry-based developerscan focus on specific system design issues that they want toinvestigate, without getting concerned about the low leveldetails related to Cloud-based infrastructures and services.CloudSim offers the following novel features: (i)support for modeling and simulation of large scale Cloudcomputing infrastructure, including data centers on a singlephysical computing node; and (ii) a self-contained platformfor modeling data centers, service brokers, scheduling, andallocations policies. Among the unique features ofCloudSim, there are: (i) availability of virtualizationengine, which aids in creation and management of multiple,independent, and co-hosted virtualized services on a datacenter node; and (ii) flexibility to switch between spacesharedand time-shared allocation of processing cores tovirtualized services. These compelling features ofCloudSim would speed up the development of newresource allocation policies and scheduling algorithms forCloud computing.
2. Key Concepts and Terminologies
This section presents background information on variousarchitectural elements that form the basis for Cloudcomputing. It also presents requirements of variousapplications that need to scale across multiplegeographically distributed data centers owned by one ormore service providers. As development of resourceallocation and application scaling techniques and theirperformance evaluation under various operationalscenarios in a real Cloud environment is difficult and hardto repeat; we propose the use of simulation as an alternateapproach for achieving the same.
2.1 Cloud computing
Cloud computing can be defined as “a type of parallel anddistributed system consisting of a collection of interconnectedand virtualized computers that are dynamicallyprovisioned and presented as one or more unifiedcomputing resources based on service-level agreementsestablished through negotiation between the serviceprovider and consumers” [1]. Some examples of emergingCloud computing infrastructures are Microsoft Azure [2],Amazon EC2, Google App Engine, and Aneka [3].Emerging Cloud applications such as social networking,gaming portals, business applications, content delivery, andscientific workflows operate at the highest layer of thearchitecture. Actual usage patterns of many real-worldapplications vary with time, most of the time inunpredictable ways. These applications have differentQuality of Service (QoS) requirements depending on timecriticality and users’ interaction patterns (online/offline).
2.2 Layered Design
Figure 1 shows the layered design of service-orientedCloud computing architecture. Physical Cloud resourcesalong with core middleware capabilities form the basis fordelivering IaaS. The user-level middleware aims atproviding PaaS capabilities. The top layer focuses onapplication services (SaaS) by making use of servicesprovided by the lower layer services. PaaS/SaaS servicesare often developed and provided by 3rd party serviceproviders, who are different from IaaS providers [13].
User-Level Middleware: This layer includes the softwareframeworks such as Web 2.0 Interfaces (Ajax, IBMWorkplace) that help developers in creating rich, costeffectinguser-interfaces for browser-based applications.The layer also provides the programming environments andcomposition tools that ease the creation, deployment, andexecution of applications in Clouds.

Download full report
http://arxivpdf/0907.4878