Cloud sim & greencloud


Published on

this presentation is held in first national workshop of cloud computing by myself in Amirkabir university in 31 october and 1 november.
i hope it will be practical after u read it :)

Published in: Education, Technology, Business
1 Comment
1 Like
  • ms. neda maleki. i'm doing my thesis on power saving algorithms in cloud. I need your experience about CloudSim. i've sent an email as well. reply me please.
    Are you sure you want to  Yes  No
    Your message goes here
No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Cloud sim & greencloud

  1. 1. First National Workshop of Cloud ComputingAmirkabir University of TechnologyPersented by: Neda Malekinedamaleki87@gmail.comCloudSim: A Toolkit for Modeling andSimulation ofCloud Computing Environments
  2. 2. OutLine• Introduction• Related Work• CloudSim Architecture• CloudSim Modelings• Design and Implementation• CloudSim Steps• Conclusions and Future works• Green Cloud
  3. 3. Introduction(1/2):Cloud• Cloud computing delivers:XaaS• X:{Software, Platform, Infrastructure }So users can access anddeploy applications fromanywhere in the Internet
  4. 4. Introduction(2/2):WhySimulation?Cloud Providor Challenges:• Maintain Quality of Service• Efficient Resourse Utilization• Dynamic Workload• Violation of Service Level Agreement• Difficulties in TestingIt’s not possible to perform benchmarkingexperiments in repeatable, dependable, andscalable environment using real-world Cloud.Possible alternative: Simulation Tool
  5. 5. Related WorksGrid simulators:GridSimSimGridOptoSimGangSimBut none of them areable to isolate themulti-layer serviceabstractions(SaaS/PaaS/IaaS)differentiation andmodel the virtualizedresources required byCloud. So:
  6. 6. Main Contribution:CloudSim A holistic software framework formodeling Cloud computing environmentsAndPerformance testing application services.
  7. 7. Features & AdvantagesFeatures• Discrete Time Event-Driven• Support modeling and simulation of large scaleCloud computing environments, including datacenters• Support simulation of network connections amongsimulated elementsAdvantages• Time effectiveness• Flexibility and applicability• Test policies in repeatable and controllableenvironment• Tune system bottlenecks before deploying on realclouds
  8. 8. Layered CloudSim Architecture(1/7)
  9. 9. Modeling in Cloudsim (1/5) Modeling DataCenter Modeling VM Allocation Modeling Network Behavior Modeling Dynamic Workloads Modeling Power Consumption
  10. 10. CloudSim Steps(1/2)abroker(VMs , Apps)CloudInformationService(CIS)Is Registered allDatacenters andtheircharacteristicsCloudDatacenter ACloudDatacenter BCloudDatacenter C
  11. 11. Allocation Policies: EnoughCapacity,Ram,Storage,BandwidthVM1,V10,VM6,VM7VM2,VM4VM9,V3,VM5VM8Scheduling Policies: Sharing of Host Mipsbetween VMs• Space Shared• Time Shared
  12. 12. DataCenter Modeling Number of Hosts, VMs and Cloudlets (tasks)o Host(mips, ram, storage, bandwidth)o Datacenter(arch, os, vmm, hostlist, costmem/bw/storage) VMo MIPS, pesNumber(no. ofcpu), Ram(MB), BW(MB/s) Cloudleto Length (MI), pesNumber, input Size, output
  13. 13. VM Allocation Modeling• Time Shared policy• Space Shared Policy
  14. 14. Simulation Setup:========== OUTPUT ==========Cloudlet ID STATUS Data center ID VM ID Time StartTime Finish Time0 SUCCESS 2 0 20.1 2.12 SUCCESS 2 0 20.1 2.11 SUCCESS 2 1 20.1 2.13 SUCCESS 2 1 20.1 2.1*****Datacenter: Datacenter_0***** 1 datecenter 1 dual-core host, each coremips: 1000 2 vm, mips:1000 4 cloudlets, length: 1000mips core1 deal with two cloudlets(t1 and t2), and core2 deal withthe other two cloudlets(t3 and t4), so, all cloudlets shouldfinished at 2.1s
  15. 15. Network Modeling• Latency MatrixDelay time from entity i toentity jEntity i Entity j
  16. 16. Dynamic Workload Modeling• The Strategy is to Vary VM Utilization!25% 43% 60% 30% 10% 90% ….Delay= not all thetime, CPU is utilized
  17. 17. Design and Implementation(1/2)CloudSim Class Design Diagram
  18. 18. Design and Implementation(2/2)Simulation Data Flow
  19. 19. Design and Impelementation(3/4)CloudSim Sequence Diagram
  20. 20. Conclusion Time effectiveness Flexibility and applicability Test services in repeatable andcontrollable environment Tune system bottlenecks beforedeploying on real clouds
  21. 21. Green Cloud
  22. 22. Power(1/4):Powering CloudInfrastructure• Modern data centers, operating under theCloud computing model, are hosting a varietyof applications ranging from those that run fora few seconds (e.g. serving requests of webapplications such as e-commerce and socialnetworks portals) to those that run for longerperiods of time (e.g. large datasetprocessing).• So, Cloud Data Centers consume excessiveamount of energy:• According to McKinsey report on “RevolutionizingData Center Energy Efficiency” :• A typical data center consumes as much energy as25,000 households!!!
  23. 23. Power (1/2) Data centers are not onlyexpensive to maintain, butalso unfriendly to theenvironment. High energy costs and hugecarbon emission are incurreddue to the massive amount ofelectricity needed to power andcool the numerous servershosted in these data centers.
  24. 24. Power Consumption in the DatacenterCompute resourcesand particularly serversare at the heart of acomplex, evolvingsystem! TheyConsumes most power.Where Does the Go?Google Datacenter2007Power
  25. 25. Levels of PowerConsideration(1/2):System levelSystem levelDPMsDVSDPSDVFSDCDSPMsLow Level Design:Gates,Transistor The objective of PA computing/communications is to improvepower management and consumption using the awareness ofpower consumption of devices. Recent devices (CPU, disk, communication links, etc.) supportmultiple power modes.
  26. 26. DVS(Dynamic Voltage Scaling)• DVS (Dynamic Voltage Scaling) technique– Reducing the dynamic energy consumption by lowering the supply voltage at thecost of performance degradation– Recent processors support such ability to adjust the supply voltage dynamically.– The dynamic energy consumption = * Vdd2 * f• Vdd : the supply voltage• f : the number of clock cycle• An example5.0210ms 25msdeadlinepowerpower deadline10ms 25ms(a) Supply voltage = 5.0 V (b) Supply voltage = 2.0 V2.02
  27. 27. Levels of PowerConsideration(2/2):DataCenter LevelData center levelVirtualizationSystem resourcesTarget systemsGoalPower saving techniquesWorkloadYesNoMultiple resourcesSingle resourceHomogeneousHeterogeneousMinimize power / energyconsumptionMinimize performancelossDVFSMeet power budgetResource throttlingDCDArbitraryReal-time applicationsHPC-applicationsWorkload consolidation
  28. 28. A Key to Power Saving!Power On Power OffPool ofphysicalcomputernodesVirtualization layer(VMMs, local resources managers)Consumer, scientific and businessapplicationsGlobal resource managersUser User UserVM provisioning SLA negotiation Application requestsVirtualMachinesandusers’applications
  29. 29. WWW: Three Sub Problems• When to migrate VMs?• Host overload detection algorithms• Host underload detection algorithms• Which VMs to migrate?• VM selection algorithms• Where to migrate VMs?• VM placement algorithms
  30. 30. Algorithms in each w Host overload detection Adaptive utilization threshold based algorithms Median Absolute Deviation algorithm (MAD) Interquartile Range algorithm (IQR) Regression based algorithms• Local Regression algorithm (LR)• Robust Local Regression algorithm (LRR) Host underload detection algorithms Migrating the VMs from the least utilized host VM selection algorithms Minimum Migration Time policy (MMT) Random Selection policy (RS) Maximum Correlation policy (MC) VM placement algorithms Heuristic for the bin-packing problem – Power-Aware Best FitDecreasing algorithm (PABFD)
  31. 31. Performance MetricsSLA violation metrics• Overloading Time Fraction (OTF) - the timefraction, during which active hosts experiencedthe 100% CPU utilization• Performance Degradation due to VM Migrations(PDM)• A combined SLA Violation metric (SLAV):SLAV = OTF * PDMA combined metric that captures both energyconsumption and the level of SLAviolations, Energy and SLA Violation (ESV):ESV = Energy * SLAV
  32. 32. Real Workloads• Workload traces from more than 1000 VMs fromservers located in more than 500 places around theworld.• The data were obtained from the CoMon project, amonitoring infrastructure for PlanetLab• PlanetLab is a distributed execution environment fordoing benchmarked experiments . Totally it is aglobal research network that supports thedevelopment of new network services.• A Data Center consisting 800 heterogeneousphysical servers containing HP ProLiant ML110 G4and HP ProLiant ML110 G5 servers.• More than 1000 Heterogeneous VMs correspondingto Amazon EC2 instance types.
  33. 33. Content of WorkLoad Files These files contain CPU utilization values measuredevery 5 minutes in PlanetLabs VMs for one day so:One day=24 hours= 5minutes*288 CloudSim contain a class called :UtilizationModelPlanetLabInMemorywhich can be used to read those workload traces. An example: String inputFolder =Dvfs.class.getClassLoader().getResource("workload/planetlab").getPath(); String outputFolder = "output"; String workload = "20110303"; // PlanetLab workloadNumber ofSamples
  34. 34. References R. Buyya, A. Beloglazov, J.Abawajy, Energy-Efficient Management ofData Center Resources for CloudComputing: A Vision, ArchitecturalElements, and OpenChallenges, Proceedings of the 2010International Conference on Parallel andDistributed Processing Techniques andApplications (PDPTA2010), LasVegas, USA, July 12-15, 2010. A. Beloglazov, R. Buyya, Y. Lee, A.Zomaya, A Taxonomy and Survey ofEnergy-Efficient Data Centers and CloudComputing Systems, Advances inComputers, Volume 82, 47-111pp, M.Zelkowitz(editor), Elsevier, Amsterdam, TheNetherlands,March 2011. S. Garg, C. Yeo, A Anandasivam, R.Buyya, Environment-ConsciousScheduling of HPC Applications onDistributed Cloud-oriented Data
  35. 35. Thanks for your attention!Any Questions , Suggestions andComments?