• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Energy-aware VM Allocation on An Opportunistic Cloud Infrastructure
 

Energy-aware VM Allocation on An Opportunistic Cloud Infrastructure

on

  • 461 views

UnaCloud is an opportunistic based cloud infrastructure ...

UnaCloud is an opportunistic based cloud infrastructure
(IaaS) that allows to access on-demand computing
capabilities using commodity desktops. Although UnaCloud
tried to maximize the use of idle resources to deploy virtual
machines on them, it does not use energy-efficient resource
allocation algorithms. In this paper, we design and implement
different energy-aware techniques to operate in an energyefficient
way and at the same time guarantee the performance
to the users. Performance tests with different algorithms and
scenarios using real trace workloads from UnaCloud, show how
different policies can change the energy consumption patterns
and reduce the energy consumption in opportunistic cloud
infrastructures. The results show that some algorithms can
reduce the energy-consumption power up to 30% over the
percentage earned by opportunistic environment.

Statistics

Views

Total Views
461
Views on SlideShare
458
Embed Views
3

Actions

Likes
0
Downloads
7
Comments
0

1 Embed 3

https://twitter.com 3

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    Energy-aware VM Allocation on An Opportunistic Cloud Infrastructure Energy-aware VM Allocation on An Opportunistic Cloud Infrastructure Presentation Transcript

    • Energy-aware VM Allocation on An Opportunistic Cloud InfrastructureExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The NetherlandsHarold Castro, Mario Villamizar(@mariocloud)Department of Systems and ComputingEngineeringUniversidad de los AndesColombiaCesar Díaz, Johnatan Pecero,Pascal BouvryComputer Science and CommunicationsResearch UnitUniversity of LuxembourgLuxembourg
    • THE PROBLEMENERGY IN AN OPPORTUNISTIC CLOUD ENVIRONMENTENERGY-AWARE TECHNIQUESALGORITHM TO CALCULATE THE ECREXPERIMENTAL RESULTSCONCLUSIONS AND FUTURE WORKExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • THE PROBLEMMore than 2000 CPU coresExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • HOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • HOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • Ubuntuwith OGEHOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • Ubuntuwith OGEVMs begin to processjobs of thebioinformatics clusterHOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • HOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • Debianwith PBSHOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • Debianwith PBSUbuntuwith OGEBoth clusters are being executedon the same physical/sharedcommodity infrastructure.HOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • HOW UNACLOUD WORKSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • THE PROBLEMVM1 VM2 VM3 VM48 CPU Cores 8 CPU Cores 8 CPU Cores 8 CPU Cores2 CPU Cores 2 CPU Cores 2 CPU Cores 2 CPU CoresHow to assign the Set of Virtual Machines (SVM) to the Set of PhysicalMachines (SPM) trying to reduce the Energy Consumption?Set of Virtual Machines (SVM)Set of Physical Machines (SPM)4. TURNED OFF2. BUSY1. IDLE0% 10% 0% 0%3. HIBERNATEDExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • ENERGY IN AN OPPORTUNISTIC CLOUD ENVIRONMENTT = Execution Time of a Taskf(x) = Function that return the ECR of a PM given its CPU usedECRmon = ECR consumed by the monitor of the PMECRuser = ECR consumed by the user using a PMECRhib = ECR consumed by the PM whet it is hibernatedExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • ENERGY IN AN OPPORTUNISTIC CLOUD ENVIRONMENTf(x) = Function that return the ECR of a PM given its CPU usedExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • ENERGY IN AN OPPORTUNISTIC CLOUD ENVIRONMENTThe best desktop machines to deploy a VM are those being used(state 2), followed by the machines that are on idle (state 1),hibernation (state 3) and turned off (state 4) state.UnaCloudExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • A. Minimize the ECR using aminimum set of PhysicalMachines preferably in busystate; increasing theExecution Time of themetatask (SVM) and executingseveral VMs on a PM.ENERGY-AWARE TECHNIQUESB. Assign to each PM(preferably in idle, hibernatedor turned off state) only oneVM, avoiding the penaltycaused by the execution ofmultiple VMs on the same PM,and increasing the ECR due tomore PMs are in execution.REDUCE ECR RESULTS FASTERExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • We propose three resource allocation algorithms that consider thefollowing assumptions and goals: Reduce the ECR required to execute a metatask (SVMs). Several VMs can be executed on a PM. The assignation of VMs to PMs is made in batch mode. There are enough CPU cores to supports the CPU requirementsof the SVM. It is better to assign a VM to a PM with a user (busy state) or to aPM (with available cores) that is already executing another VM(busy state). Physical machines are homogeneous.ENERGY-AWARE TECHNIQUESExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • VM1 VM2 VM3 VM48 CPU Cores 8 CPU Cores 8 CPU Cores 8 CPU Cores2 CPU Cores 2 CPU Cores 2 CPU Cores 2 CPU CoresThree different strategies/techniques were defined and tested to reducethe ECR consumed by UnaCloudSet of Virtual Machines (SVM)Set of Physical Machines (SPM)4. TURNED OFF2. BUSY1. IDLE0% 10% 0% 0%3. HIBERNATEDENERGY-AWARE TECHNIQUESExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • 1. CUSTOM ROUND ROBIN ALLOCATIONSVM in arbitrary order• It does not consider the state of the PM.• It does not use a minimum SPM.• VMs with low requirements may be initially assigned to PMs withlarge CPU capabilities (blocking VMs with larger requirements).SPM in arbitrary orderExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • 2. SORTING VMS AND PMs TO MINIMIZE THE USE OF PMsSVM descending ordered by cores required and execution time.SPM ordered by (1) PMs running VMs, (2) PMs in busy state and (3)PMs with more available CPU cores.ExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • 2. SORTING VMS AND PMs TO MINIMIZE THE USE OF PMs We identified that if VMs withsimilar execution times areexecuted on the same PMsthe ECR required to executethe SVM is lower. Every PMs used to executedthe SVM will execute tasks atpeak capacity during asimilar period of time.PM1 PM28 Cores 8 CoresVM14 Cores10 HoursVM24 Cores1 HourVM34 Cores10.5 HoursPM1 PM28 Cores 8 CoresVM14 Cores10 HoursVM24 Cores1 HourVM34 Cores10.5 HoursWith Algorithm 2ExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    •  We identified that if VMs withsimilar execution times areexecuted on the same PMsthe ECR required to executethe SVM is lower. Every PMs used to executedthe SVM will execute tasks atpeak capacity during asimilar period of time.PM1 PM28 Cores 8 CoresVM14 Cores10 HoursVM24 Cores1 HourVM34 Cores10.5 HoursPM1 PM28 Cores 8 CoresVM14 Cores10 HoursVM24 Cores1 HourVM34 Cores10.5 HoursWith Algorithm 33. EXECUTING VMS WITH SIMILAR EXECUTION TIMEExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • 3. EXECUTING VMS WITH SIMILAR EXECUTION TIMEExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • ALGORITHM TO CALCULATE THE ECR This algorithm is used once thescheduling process (using one ofthe 3 previous algorithms) hasfinished. The ECR consumed by every PMis calculated based on the ECR ofeach VM. The goal of this algorithm is tocalculate the total powerconsumed by the VMs.ExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    •  We conduct experiments based onsimulations. We use workloads based on realVMs production traces (collectedduring one year). We consider that a givenpercentage of PMs has a userworking on it. We vary the percentage of busymachines from 10% up to 50%with increments of 10%.EXPERIMENTAL RESULTSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • EXPERIMENTAL RESULTSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • EXPERIMENTAL RESULTSExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • EXPERIMENTAL RESULTS On average Algorithm 2 andAlgorithm 3 save up to 36% of PMsmore than Algorithm 1 when only10% of the PMs are busy by users. When the number of PMs withusers increases up to 50%,Algorithm 2 and Algorithm 3require 26% less PMs than RR.ExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • EXPERIMENTAL RESULTS For the first scenario, the averagegain on ECR by Algorithm 2 is upto 19% and up to 20% byAlgorithm 3. In scenario 2, Algorithm 2 andAlgorithm 3 save up to 29% and30% more than Algorithm 1,respectively. Algorithm 3 saves 2% more energyin both scenarios than Algorithm2.ExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    •  We developed and simulateddifferent energy-aware algorithmsto allow UnaCloud to operate in anenergy efficient way. We will implement the newproposed algorithms on theUnaCloud infrastructure. We will try more allocationstrategies that consider VMsshared resources constraints thatcan lead to performancedegradation on the quality ofservice.CONCLUSIONS AND FUTURE WORKExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • THANKS FOR YOUR ATTENTION!ExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands
    • CONTACT INFORMATIONmj.villamizar24@uniandes.edu.co mjvc007@hotmail.com@mariocloudhttp://linkedin.com/in/mariojosevillamizarcanoMario José Villamizar CanoExtremeGreen: Extreme Green & Energy Efficiency in Large ScaleDistributed Systems (CCGrid 2013) – Delft, The Netherlands