Cloud Simulation for

Building and Evaluating Solutions for Cost

and Energy-Efficient Cloud Computing

Dr. Rajkumar Buyya

Cloud Computing and Distributed Systems (CLOUDS) Lab Manjrasoft
Dept. of Computer Science and Software Engineering Innovative Solutions for Cloud Computing

The University of Melbourne, Australia Dr Rajkumar Buyya

www.cloudbus.org Chief Executive Officer

Manjrasoft Pty Ltd

www.buyya.com
Room 5.31, ICT Building, 111, Barry Street, Carlton,
Melbourne, VIC 3053, Australia
P: +61-3-8344 1344 | F : +61-3-9348 1184

www.manjrasoft.com
E: raj@manjrasoft.com
http://www.manjrasoft.com

Major Sponsors/Supporters

Manjrasoft
 Outline
Manjrasoft

 Cloud Computing at a Glance

 Marketplace of Computing Utilities and Challenges
 Market-Oriented Cloud Architecture
 SLA-oriented Resource Management
 Policies for VM Provisioning, VM-to-PM Mapping,
Tasks-to-VM Mapping, Multi-Cloud Brokering
 Experimental Methodologies
 Approaches and CloudSim Toolkit
 Energy-Efficient Resource Management
 Within a Cloud Data Center
 Across Multiple Data Centers (InterCloud)
 Summary and Thoughts for Future

2
 Cloud as a Marketplace of
SaaS

PaaS

Manjrasoft
“Computing Utilities” IaaS

Compute Cloud Cluster (VM Pool)

User User Pool node
VM VM VM
Cloud Manager
Coordinator

Cloud Broker 1 Cloud Broker N Publish Offers Pool node Pool node
...... VM VM VM

VM VM
Negotiate/Bid
Request Directory
Capacity
Bank Cloud
Coordinator
Auctioneer
Storage Cloud
Cloud
Coordinator

Enterprise Cloud Exchange

Resource (CEx)
Server
(Proxy) Storage Cloud Compute Cloud

3 Enterprise IT Consumer
 Cloud Characteristics and Resource
Manjrasoft
Management Challenging Issues
 Shared, Dynamic
 Heterogeneous
 Multiple Network Topologies
 Variety of Resources
 Compute, Memory, Storage, Network, Cooling
 Varied Pricing Schemes
 Distributed and Autonomous Providers
 Independent Failures
 Variety of Applications and Programming Models
 Compute-Intensive, Data-Intensive…| SPMD, MapReduce, Workflow, …

4
 Outline
Manjrasoft

 Cloud Computing at a Glance

 Marketplace of Computing Utilities and Challenges
 Market-Oriented Cloud Architecture
 SLA-oriented Resource Management
 Policies for VM Provisioning, VM-to-PM Mapping,
Tasks-to-VM Mapping, Multi-Cloud Brokering
 Experimental Methodologies
 Approaches and CloudSim Toolkit
 Energy-Efficient Resource Management
 Within a Cloud Data Center
 Across Multiple Data Centers (InterCloud)
 Summary and Thoughts for Future

5
 Market-Oriented Cloud Architecture: QoS negotiation
and SLA-based Resource Allocation
Manjrasoft
Users/
Brokers

Service Request Examiner and

Admission Control
- Customer-driven Service Management
- Computational Risk Management
SLA - Autonomic Resource Management
Resource
Allocator Pricing Accounting

VM Service Request
Monitor Dispatcher Monitor

Virtual
Machines
(VMs)

Physical
Machines

6
 Cloud to expand (lease more resources) with
increase in demand
Manjrasoft
Users/
Brokers

Service Request Examiner and

Admission Control
- Customer-driven Service Management
- Computational Risk Management
SLA - Autonomic Resource Management
Resource
Allocator Pricing Accounting

VM Service Request
Monitor Dispatcher Monitor

Virtual
Machines
(VMs)

Multi-Clouds
Physical
Machines

7
 Cloud to shrink (unlease resources) with
decrease in demand
Manjrasoft
Users/
Brokers

Service Request Examiner and

Admission Control
- Customer-driven Service Management
- Computational Risk Management
SLA - Autonomic Resource Management
Resource
Allocator Pricing Accounting

VM Service Request
Monitor Dispatcher Monitor

Virtual
Machines
(VMs)

Physical
Machines

8
 Outline
Manjrasoft

 Cloud Computing at a Glance

 Marketplace of Computing Utilities and Challenges
 Market-Oriented Cloud Architecture
 SLA-oriented Resource Management
 Policies for VM Provisioning, VM-to-PM Mapping,
Tasks-to-VM Mapping, Multi-Cloud Brokering
 Experimental Methodologies
 Approaches and CloudSim Toolkit
 Energy-Efficient Resource Management
 Within a Cloud Data Center
 Across Multiple Data Centers (InterCloud)
 Summary and Thoughts for Future

9
 Requirements for Cost and Energy-efficient
Manjrasoft
Clouds
 Innovative Resource Management and Scheduling (RMS) Policies
 Brokering of Cloud CDC Resources
 Provisioning VMs
 Mapping VMs to PMs
 Mapping Tasks to VMs (Scheduling)
 Consolidation of VMs and Resources
 What is the “best” approach for Evaluating RMS policies?

10
 Experimental Methodologies
Manjrasoft

Model
Benchmark Simulation
Software
Real

In-situ Emulation

Real Model
Hardware

J. Gustedt, E. Jeannot, M. Quinson. “Experimental methodologies for large-scale systems:

11 a survey”, Parallel Processing Letters 19(3):399-418. World Scientific, Sep 2009.
 Experimental Methodologies
Manjrasoft

Benchmark Simulation

Model
(TPC, SPEC, (CloudSim,
Software CloudStone) NS2)

In-situ
Real
Emulation
(AWS, Azure,
(EMUSIM)
Aneka..)

Real Model
Hardware

12
 Characteristics (Pros/Cons) of Experimental
Manjrasoft
Methodologies

In-Situ Benchmark Emulation Simulation

Setup Cost High Medium Medium Very Low/Nil
Control Low Low High Very High
Repeatability Low High Medium Very High
Heterogeneity Low/fixed Medium Low/fixed Configurable
Scale Low/Medium Medium Low Configurable
Prominent Users Industry Academia Designers Academia
Execution Time Slow Medium Very Slow Fast

13
 Aneka: Cloud Application Platform (CAP) for Resource-
Intensive/Elastic Apps
Manjrasoft

 World-first platform Multiple Applications

supporting multiple Cloud
programming models (Task, Thread Task ... MapReduce
Thread, MapReduce)
 SDK (Software 1. SDK
Development Kit) containing
APIs for multiple programming
models and tools
Aneka
 Runtime Environment for
managing application
execution on Clouds
 Suitable for 2. Runtime
 Development of Enterprise
Cloud Applications Multi-core Cluster Grid Cloud
 Cloud enabling legacy 2100 2100 2100 2100

applications
 Portability for Customer Apps: 2100 2100 2100 2100

 Enterprise ↔ Public Clouds

 .NET/Win ↔ Mono/Linux
14 Multiple Infrastructures
 CloudSim Overview
Manjrasoft

15
 CloudSim elements
C1 C2 C3 CN
Manjrasoft
Scheduler Scheduler

VM1 VM2 VM3 VM4

VMM VMM

PEs PEs

Hosts Hosts

Data Center 1 Data Center 2

Network

VM1 ... VMX

User/Broker
C1 C2 ... CN
16
 CloudSim Architecture
Manjrasoft

User code
Simulation Cloud User Application SaaS
Specification Scenario Requirements … Configuration

Scheduling
User or Data Center Broker
Policy

CloudSim
User
Virtual
Interface Cloudlet
Machine
Structures PaaS
VM Cloudlet VM
Services Execution Management

Cloud VM CPU Memory Storage Bandwidth

Services Provisioning Allocation Allocation Allocation Allocation

Cloud Events Cloud

Sensor Data Center
Resources Handling Coordinator

IaaS
Network Message delay
Network Topology Calculation

CloudSim core simulation engine

17
 VM Provisioning
Manjrasoft

 VMAllocationPolicy
●
Management of VMs C1 C2

(creation...): OpenStack Alloc

●
Defined @ Data Center level VM1 VM2
–DifferentData Centers in the same VMM
simulation may use different policies
PEs
●
Which host will receive the
VM? Hosts
–Loadbalance
–Consolidation (green IT)
Data Center

●
Migration

18
 VM scheduling
Manjrasoft
 VMScheduler
●
Defined in Host level C1 C2
–Different hosts in the same
Alloc
Data Center may have
VM1 VM2
different policies
●
How to share PEs among VMM

VMs in the host? PEs

–Xen: RR, Credit
Hosts
–Time-shared, space-shared,
proportional Data Center
●
Provisioning policies for
resources (memory and
bandwidth) also
19
customizable
 Task scheduling
Manjrasoft

 CloudletScheduler
●
Defined in VM level C1 C2
–Different VMs in the same Alloc

host may have different VM1 VM2

policies VMM
●
How to share processing PEs

power allocated to a VM Hosts

among Cloudlets?
–OS scheduling Data Center

–Time-shared and space-

shared
20
 Broker
C1 C2
Manjrasoft
VMScheduler
 DatacenterBroker
VM1 VM2
●
“Cloud scheduling”
VMM
–Selects the Data Center to
PEs
host VMs
–Selects Hosts
VMs to run
Cloudlets Data Center

–Application
model: PS?
Network
BoT? Workflow?
–Economic decisions VM1 VM2
C1
C2
21 User/Broker
 Outline
Manjrasoft

 Cloud Computing at a Glance

 Marketplace of Computing Utilities and Challenges
 Market-Oriented Cloud Architecture
 SLA-oriented Resource Management
 Policies for VM Provisioning, VM-to-PM Mapping,
Tasks-to-VM Mapping, Multi-Cloud Brokering
 Experimental Methodologies
 Approaches and CloudSim Toolkit
 Energy-Efficient Resource Management
 Within a Cloud Data Center
 Across Multiple Data Centers (InterCloud)
 Summary and Thoughts for Future

22
 Key Issue: Sustainability
Global Cloud Electricity Consumption
Manjrasoft

1
91 billion kWh
in the U.S. =
Cloud Data Centers(DCs)
X2
Twice of
electricity consumption
in 2013 in New York City

 If the Cloud were a country, it would have the 5th largest electricity demand in the world.  high carbon
footprint.

23
Source: Natural Resources Defense Council / Greenpeace International / ACS
 Where Does the Power Go?
Manjrasoft

Power Consumption in the Datacenter Server/Storage 50%

Computer Rm. AC 34%

Network 7%

Conversion 7%

Lighting 2%

Compute resources and

particularly servers are at
the heart of a complex,
evolving system!

Source: APC
24
 Case Study: Dynamic VM Consolidation
Manjrasoft

User User User

VM provisioning SLA negotiation Application requests

Global resource managers

Virtual Consumer, scientific and business
Machines applications
and
users’
applications

Virtualization layer
(VMMs, local resources managers)

Pool of
physical
computer
27 nodes Power On Power Off
 Consolidation/Overbooking
Manjrasoft

Wasted

28
 Three Sub-Problems
Manjrasoft

 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

29
 Proposed “Power-Aware” Algorithms
Manjrasoft
• Host overload detection
• Adaptive utilization threshold based algorithms
• Median Absolute Deviation (MAD) algorithm
• Interquartile Range (IQR) algorithm
• Regression based algorithms
• Local Regression (LR) algorithm
• Robust Local Regression (LRR) algorithm
• Host underload detection algorithms
• Migrating the VMs from the least utilized host
• VM selection algorithms
• Minimum Migration Time (MMT) policy
• Random Selection (RS) policy
• Maximum Correlation (MC) policy
• VM placement algorithms
• Heuristic for the bin-packing problem – Power-Aware Best Fit
Decreasing (PABFD) algorithm
30
 Simulation Setup
Manjrasoft

• CloudSim with a power package

• A Data Center consisting:
• 800 heterogeneous physical servers containing HP
ProLiant ML110 G4 and HP ProLiant ML110 G5
servers.
• More than 1000 Heterogeneous VMs
corresponding to Amazon EC2 instance types
 Workload traces from more than 1000 VMs from
servers located in more than 500 places around
the world.
 The data were obtained from the CoMon project, a
monitoring infrastructure for PlanetLab
32
 Best Algorithm Combinations and Benchmark
Manjrasoft
Algorithms

Dynamic VM consolidation significantly reduces energy consumption compared to non-power aware

allocation and static allocation policies, like DVFS, NPA (non-power aware)

33
 OpenStack Neat: EE VM Consolidation
Manjrasoft

34
 Outline
Manjrasoft

 Cloud Computing at a Glance

 Marketplace of Computing Utilities and Challenges
 Market-Oriented Cloud Architecture
 SLA-oriented Resource Management
 Policies for VM Provisioning, VM-to-PM Mapping,
Tasks-to-VM Mapping, Multi-Cloud Brokering
 Experimental Methodologies
 Approaches and CloudSim Toolkit
 Energy-Efficient Resource Management
 Within a Cloud Data Center
 Across Multiple Data Centers (InterCloud)
 Summary and Thoughts for Future

35
 Conclusions
Manjrasoft
●
Simulations enable creation of scalable, repeatable, and
controllable experimental environments
●
CloudSim is a tool for simulation of Cloud computing
environments for evaluation of RMS algorithms/policies:
–Provisioning of resources
–Scheduling of applications

●
RMS polices focused on:
–Cost Minimization
–Energy-efficiency (Min energy consumption and Carbon footprint)
–Reliability

●
Once proven in CloudSim, you can incorporate them or
plug-in for “Real” software systems like OpenStack
●
CloudSIm has 3000+ users world-wide.

36
 Plugins/Other Efforts
Manjrasoft

●
Power-aware computing modules (INRIA France)
●
Software-defined networks (CloudSimSDN)
●
Containers
●
Workflows (WorkflowSim by University of Southern California)
●
Modeling of tight-coupled applications (like MPI)
●
Big Data Simulations (CloudSimMapReduce)
●
CloudAuctions
●
Reliability modelling (CloudSimFT)
●
Linking CloudSim to OpenStack (CloudSimSpot)
●
Fog and Computing Simulations (iFogSim)

37
 iFogSim Toolkit for Simulation of Fog and
Manjrasoft
Edge Computing Environments

39
 Thanks for your attention!
Manjrasoft

 Are there any

 Questions?
 Comments/Suggestions
Manjrasoft

We welcome you to:

Study/Research with Us | Do Business with us!
http:/www.cloudbus.org | www.Manjrasoft.com
rbuyya@unimelb.edu.au | raj@manjrasoft.com
40
 References Papers
Manjrasoft

 Rodrigo N. Calheiros, Rajiv Ranjan, Anton Beloglazov, Cesar A. F. De Rose, and Rajkumar
Buyya, CloudSim: A Toolkit for Modeling and Simulation of Cloud Computing Environments
and Evaluation of Resource Provisioning Algorithms, Software: Practice and Experience,
41(1): 23-50, Wiley Press, New York, USA, January 2011.
 A. Beloglazov and R. Buyya, Managing Overloaded Hosts for Dynamic Consolidation of Virtual
Machines in Cloud Data Centers Under Quality of Service Constraints, IEEE Transactions on
Parallel and Distributed Systems (TPDS), 24(7): 1366-1379, IEEE CS Press, USA, July 2013.
 Anton Beloglazov and Rajkumar Buyya, OpenStack Neat: A Framework for Dynamic and
Energy-Efficient Consolidation of Virtual Machines in OpenStack Clouds, Concurrency and
Computation: Practice and Experience (CCPE), 27(5): 1310-1333, Wiley Press, New York,
USA, April 2015.
 Harshit Gupta, Amir Vahid Dastjerdi , Soumya K. Ghosh, and Rajkumar Buyya, iFogSim: A
Toolkit for Modeling and Simulation of Resource Management Techniques in Internet of
Things, Edge and Fog Computing Environments, Software: Practice and Experience (SPE),
47(9):1275-1296, Wiley Press, New York, USA, September 2017.

41
 A Reference Book
Manjrasoft

Morgan Kauffman, USA McGraw Hill, India China Machine Press, China

42
Manjrasoft

Manjrasoft

43
Green Cloud Computing
 Outline
Manjrasoft

 Cloud Computing at a Glance

 Cloud Benefits and Challenges
 Powering Cloud Infrastructure
 Energy Consumption, Costs, Implications
 Power-Aware Computing
 Trends, Foundations, Issues, Taxonomy
 Green Cloud Computing: Framework
 Energy-Efficient Resource Management
 Within a Cloud Data Center
 Across Multiple Data Centers (InterCloud)
 Summary and Thoughts for Future

44
 Green Cloud or Brown Cloud?
Manjrasoft

• Ideally, for every server virtualized, save

– ~$700 and ~7,000 kWh / year
– 4 tons of CO2 emissions / year
• Plus
– Power down underutilized physical servers, saving 40%
– Desktop management, saving 35% / year
• But currently
Cloud Location Estimated power % of Dirty % of Renewable
datacenters usage Energy Electricity
Effectiveness Generation
Google Lenoir 1.21 50.5% Coal, 3.8%
38.7% Nuclear
Apple Apple, NC 50.5% Coal, 3.8%
38.7% Nuclear
Microsoft Chicago, IL 1.22 72.8% Coal, 1.1%
22.3% Nuclear
Yahoo La Vista, NE 1.16 73.1% Coal, 7%
14.6% Nuclear
45
 Some Observations
Manjrasoft

 Datacenters have heterogeneous properties

– Geographically distributed datacenters (different
environmental factors and electricity prices)
– Each resource site has different CPU configurations
– Each site has different energy efficiency
– Different Carbon-footprint

46
Source: Best Practices for Data Centers: Lessons Learned from Benchmarking 22 Data Centers by Lawrence Berkeley National Laboratory’s report 46
 Green Cloud Architecture
Manjrasoft
End User

d) Allocate
service
Private
a) Request Cloud
a cloud
service

Green Broker c) Request

energy
efficiency
information

b) Request Carbon Emission

any green
Routers offer Directory
Internet
e) Request
service
allocation

Green Offer
Public Cloud A Directory
47 Public Cloud B
 Third Party: Green Offer and Carbon
Manjrasoft
Emission Directory
 Carbon Emission Directory
 Contains data on Power Usage Effectiveness (PUE), cooling
efficiency, carbon footprint, network cost
 Helps user to select cloud services with minimum carbon
footprint
 Incentive for providers
 Advertising tool to increase the market share, e.g. Google
 Require more carbon transparency from providers
 Government role by enforcing policies such as Carbon Tax
 Green Offer Directory
 Incentive for users
 Choosing more carbon efficient hours
 Lists services with their discounted prices and green hours
48
 User: Green Broker
Manjrasoft

• A typical Cloud broker

– Lease Cloud services
Schedule applications
User
–

Green Broker
Cloud Request Services
QoS
Application
Profiling
Cloud
Offers
• Green Broker
CO2 Analysis Services
– 1st layer: Analyze user
Cost
Calculator
CO2 Emission
Calculator
Green
Information requirements
System

Brokering Services such as

– 2nd layer: Calculates cost
scheduling, monitoring and carbon footprint of
Green
Policies
Cloud
Leasing
Scheduler services
– 3rd layer: Carbon aware
scheduling
Private Cloud Public Cloud

49
 Provider: Green Middleware
Manjrasoft

50
 Case Study: IaaS Cloud
Manjrasoft

 Carbon Emission Directory: Stores all carbon

emission rates for each IaaS provider

 Green Offer Directory: Receives number of VMs

that can be initiated at a particular time for
maximum energy efficiency

 Green Broker: Computes schedule with the

lowest carbon emission based on application
requirements
51
 Carbon Efficient Green Policy (CEGP)
Manjrasoft

 Collect resource requests from users and

resource site information such as VMs, carbon
emission rate, DCiE, CPU power efficiency
 Sort jobs based on deadline
 Sort resource sites based on carbon footprint:

Carbon Datacenter Energy

Emission Efficiency Efficiency of VM

 Schedule greedily the most urgent deadline jobs

on the most power efficient resource site.

52
 Simulation Setup
Manjrasoft

 Parallel Workload: first week of LLNL Thunder

trace from Parallel Workload Archive (PWA)
 Deadline generated based methodology proposed by
Irwin et al. (2004)1
 Configuration of Cloud resource sites2:

1
D. Irwin, L. Grit, and J. Chase, “Balancing risk and reward in a market-based task service,” in Proc. of the 13th IEEE International Symposium on High
Performance Distributed Computing, Honolulu, USA, 2004.
53 2
L. Wang and Y. Lu, “Efficient Power Management of Heterogeneous Soft Real-Time Clusters,” in Proc. of the 2008 Real-Time Systems Symposium,
 EDF: Carbon-Efficient Green Policy (CEGP) Vs Early Start-Time
(EST) algorithm
Manjrasoft

54
 Case 2 (InterCloud): Summary
Manjrasoft

• Presented a Carbon Aware Green Cloud Framework to

improve the carbon footprint of Cloud computing.
• Proposed framework provides incentives to both users
and providers to utilize and deliver the most “Green"
services.
• Proposed a Carbon Efficient Green Policy (CEGP) for
IaaS providers.
• Green Policy CEGP can save up to 23% energy while
reducing the carbon footprint by about 25%.

55
 Final Results
Manjrasoft

56
 Benefits of (Public) Clouds
Manjrasoft

 No upfront infrastructure investment

 No procuring hardware, setup, hosting, power, etc..
 On demand access
 Lease what you need and when you need..
 Efficient Resource Allocation
 Globally shared infrastructure, can always be kept busy by serving users from different time
zones/regions...
 Nice Pricing
 Based on Usage, QoS, Supply and Demand, Loyalty, …
 Application Acceleration
 Parallelism for large-scale data analysis, what-if scenarios studies…
 Highly Availability, Scalable, and Energy Efficient
 Supports Creation of 3rd Party Services & Seamless offering
 Builds on infrastructure and follows similar Business model as Cloud

57
Manjrasoft

 OpenStack and CloudSim

58
 A system for VM consolidation
Manjrasoft

59
 OpenStack Neat: EE VM Consolidation
Manjrasoft

60
 A sequence diagram of handling an
Manjrasoft
underload request

61