ENERGY-EFFICIENT SCHEDULING

OF HPC IN CLOUD COMPUTING

Presented by
P. Dhanraj
2009115018
Cloud Computing Definition
 Provisioning of dynamically scalable and
virtualized resources as a service over the
Internet.


◦ Multi-tenancy
◦ Device & Location independence
◦ Ability to obtain virtual computing resource on demand
◦ Provides the Illusion of infinite computing resources
◦ Self-Provisioning of virtual resources
◦ Provides the ability to pay as you go for use of
computing resources
◦ Reliability, Scalability, Security, Manageability

Types of Clouds
Hybrid Public Cloud
Cloud

Providers
Service Users
Private Cloud

Service
Cloud End-User Services (SaaS)

Cloud Platform Services (PaaS)

Cloud Providers
Cloud Infrastructure Services (IaaS)

Physical Infrastructure
Virtualization
 Virtual workspaces:
◦ An abstraction of an execution environment that can
be made dynamically available to authorised
clients by using well-defined protocols,
◦ Resource quota (e.g. CPU, memory share),
◦ Software configuration (e.g. O/S, provided services).
 Implement on Virtual Machines (VMs):
◦ Abstraction of a physical host machine,
◦ Hypervisor intercepts and emulates instructions from
VMs, and allows management of VMs,App App App
◦ VMWare, Xen, etc. OS OS OS
 Provide infrastructure API: Hypervisor
◦ Plug-ins to hardware/support structuresHardware
Virtualized Stack
Virtual Machines
 VM technology allows multiple virtual
machines to run on a single physical
machine.
App App App App App
Xen
Guest OS Guest OS Guest OS
(Linux) (NetBSD) (Windows)
VMWare
VM VM VM
UML
Virtual Machine Monitor (VMM) / Hypervisor
Denali
Hardware
etc.

Performance: Para-virtualization (e.g. Xen) is very close to raw physical performance!

ADVANTAGES
 Lower computer costs
 Improved performance
 Reduced software costs
 Instant software updates
 Improved document format compatibility
 Unlimited storage capacity
 Increased data reliability
 Easier group collaboration
 Device independence

6
DISADVANTAGES
 Requires a constant internet connection
 Features might be limited
 Can be slow
 Stored data might not be secured
 Stored data can be lost

7
Project Goal
 Energy consumption
 Profit
 Environmentally sustainable
 Workload
 CPU power efficiency
 Location
 Management system
 Architectural design
Related Works
 Bradley et al. proposed algorithms to minimize the power utilization
by using workload history and predicting future workload within
acceptable reliability.[1]


 Chen et al. simulated a cluster using both predictive queuing models

and feedback controllers to derive frequency adjustments for web
serving clusters over control intervals lasting several minutes.[2]


 Nathuji and Schwan integrated power management mechanisms and

policies with virtualization technologies to reduce power
consumption for web workloads. [3]


 Cardosa et al. exploited the minimum-maximum resource partitions

and shares features in Virtual Machine (VM) technology to manage
power in a data center.[4]



Meta-scheduling Model
 Cloud meta-scheduler
 Service requirements
 Quality of Service(QoS) requirements
 Carbon emissions can be reduced
 Data centers depending on regional
constraints
 Updating the meta-scheduler
Cloud meta-scheduling
protocol
Minimizing carbon emission
and maximizing profit(MCE-
MP)
 Find the cloud site with minimum carbon
emission
 Among these, find the maximum profit cloud
site
 Update the available slots from cloud sites
 Do steps 1 to 3 until all applications are
mapped

Time optimization policy

 Within the available budget

 Available budget
 BudgetperHour=(total budget)/(Deadline-currentTime)
 Hiring resources from Iaas
 List of available servers
 Request send to the IaaS provider
 Hired resources at the end of execution



Implementation
Cost optimization policy
Experimental scenarios

 Effect of urgency and arrival rate of

applications
 Effect of mapping policy
 Impact of carbon emission rate
 Impact of data center efficiency
 Impact of electricity cost


Analysis of results
Effect of urgency and arrival time
Effect of mapping policy
Impact of electricity cost
Conclusion and Future work

 High amount of electricity
 Resource allocation across multiple locations
 Increase the computational capacity of local
resources
 Minimizes completion time and cost
 Several IaaS provider with different prices
for their resources
References
[1] Y. Chen, A. Das, W. Qin, A. Sivasubramaniam, Q. Wang, N.
Gautam, Managing server energy and operational costs in
hosting centers, ACM SIGMETRICS Performance Evaluation
Review 33 (1) (2005) 303-314.
[2] B. Lawson, E. Smirni, Power-aware resource allocation in high-

end systems via online simulation, in: Proc. of the 19th annual
international conference on Supercomputing, Cambridge, USA,
2005, pp. 229-238.
[3] R. Nathuji, K. Schwan, Virtualpower: coordinated power

management in virtualized enterprise systems, in: SOSP '07:

Proceedings of 21st ACM SIGOPS31symposium on Operating
systems principles, ACM, New York, NY, USA, 2007, pp. 265{278.
[4] A. Verma, P. Ahuja, A. Neogi, pmapper: Power and migration

cost aware application placement in virtualized systems, in:

Proc. of the 9th ACM/IFIP/USENIX International Conference on
Middleware, Leuven, Belgium, 2008.

[5] TOP500 Supercomputers, Supercomputer's Application
Area Share, http://www.top500.org/stats/list/33/apparea
(2009).
[6] R. Buyya, C. S. Yeo, S. Venugopal, J. Broberg, I. Brandic,

Cloud Computingand Emerging IT Platforms: Vision, Hype,

and Reality for Delivering Computing as the 5th Utility,
Future Generation Computer Systems 25 (6) (2009) 599-
616.
[7] Ian Foster, What's faster-a supercomputer or EC2? ,

http://ianfoster.typepad.com/blog/2009/08/whats-fastera-supe
(Aug. 2009).
[8] R. Bianchini, R. Rajamony, Power and energy

management for server systems, Computer 37 (11)

(2004) 68-74.
[9] S. Rivoire, M. A. Shah, P. Ranganathan, C. Kozyrakis,

Joulesort: a balanced energy-e_ciency benchmark, in:

SIGMOD '07: Proceedings of the 2007 ACM.
[10] Amazon Elastic Comput cloud,

http://aws.amazon.com/ec2.

Thank you