Are Low Power Server CPUs Worth the Cost?

Slides:



Advertisements
Similar presentations
© Actility – Confidential – Under NDA 1 Advanced flexibility management: concepts and opportunities Making Things Smart.
Advertisements

Sabyasachi Ghosh Mark Redekopp Murali Annavaram Ming-Hsieh Department of EE USC KnightShift: Enhancing Energy Efficiency by.
CloudStack Scalability Testing, Development, Results, and Futures Anthony Xu Apache CloudStack contributor.
Walter Binder University of Lugano, Switzerland Niranjan Suri IHMC, Florida, USA Green Computing: Energy Consumption Optimized Service Hosting.
Power Management in Cloud Computing using Green Algorithm -Kushal Mehta COP 6087 University of Central Florida.
Energy-efficient Virtual Machine Provision Algorithms for Cloud System Ching-Chi Lin Institute of Information Science, Academia Sinica Department of Computer.
Overcoming the challenge of virtual blindness Colin Richardson on365 Ltd.
Building a Sustainable Data Center Matthew Holmes Johnson County Community College.
COMS E Cloud Computing and Data Center Networking Sambit Sahu
Energy Efficient Web Server Cluster Andrew Krioukov, Sara Alspaugh, Laura Keys, David Culler, Randy Katz.
Akhil Langer, Harshit Dokania, Laxmikant Kale, Udatta Palekar* Parallel Programming Laboratory Department of Computer Science University of Illinois at.
05/15/09 Green Data Center Program Ian Katz Power Metering, The New Frontier.
Energy Model for Multiprocess Applications Texas Tech University.
Datacenter Transformation. In the past decade US Government owned datacenters skyrocketed from 432 to 1200 However, average utilization of servers is.
Benchmarking Methodology for Virtualization Network Performance draft-huang-bmwg-virtual-network-performance-00 Lu Huang Rong Gu (Presentor) Dapeng Liu.
Lecture 2: Technology Trends and Performance Evaluation Performance definition, benchmark, summarizing performance, Amdahl’s law, and CPI.
Cloud Data Center/Storage Power Efficiency Solutions Junyao Zhang 1.
Denny Cherry Senior Database Administrator / Architect MVP, MCSA, MCDBA, MCTS, MCITP.
XI HE Computing and Information Science Rochester Institute of Technology Rochester, NY USA Rochester Institute of Technology Service.
November , 2009SERVICE COMPUTATION 2009 Analysis of Energy Efficiency in Clouds H. AbdelSalamK. Maly R. MukkamalaM. Zubair Department.
Power and Performance Modeling in a Virtualized Server System M. Pedram and I. Hwang Department of Electrical Engineering Univ. of Southern California.
Kinshuk Govil, Dan Teodosiu*, Yongqiang Huang, and Mendel Rosenblum
Michael Ikerionwu 4 th year Electronic Engineering.
Temperature Aware Load Balancing For Parallel Applications Osman Sarood Parallel Programming Lab (PPL) University of Illinois Urbana Champaign.
Tag line, tag line Power Management in Storage Systems Kaladhar Voruganti Technical Director CTO Office, Sunnyvale June 12, 2009.
Energy Savings with DVFS Reduction in CPU power Extra system power.
High Performance Computing on Virtualized Environments Ganesh Thiagarajan Fall 2014 Instructor: Yuzhe(Richard) Tang Syracuse University.
Dynamic Resource Monitoring and Allocation in a virtualized environment.
Energy Aware Consolidation for Cloud Computing Srikanaiah, Kansal, Zhao Usenix HotPower 2008.
Power-Aware Scheduling of Virtual Machines in DVFS-enabled Clusters
Energy Management in Virtualized Environments Gaurav Dhiman, Giacomo Marchetti, Raid Ayoub, Tajana Simunic Rosing (CSE-UCSD) Inside Xen Hypervisor Online.
Green Computing. Current system extremely wasteful – Need energy to power – Need energy to cool 1000 racks, 25,000 sq ft, 10MW for computing, 5 mw to.
Dana Butnariu Princeton University EDGE Lab June – September 2011 OPTIMAL SLEEPING IN DATACENTERS Joint work with Professor Mung Chiang, Ioannis Kamitsos,
Vic Liu Bob Mandeville Brooks Hickman Weiguo Hao Zu Qiang Speaker: Vic Liu China Mobile Problem Statement for VxLAN Performance Test draft-liu-nvo3-ps-vxlan-perfomance-00.
Exascale Computing. 1 Teraflops Chip Knight Corner will be manufactured with Intel’s 3-D Tri-Gate 22nm process and features more than 50 cores.
VGreen: A System for Energy Efficient Manager in Virtualized Environments G. Dhiman, G Marchetti, T Rosing ISLPED 2009.
VMWare MMU Ranjit Kolkar. Designed for efficient use of resources. ESX uses high-level resource management policies to compute a target memory allocation.
June 30 - July 2, 2009AIMS 2009 Towards Energy Efficient Change Management in A Cloud Computing Environment: A Pro-Active Approach H. AbdelSalamK. Maly.
XI HE Computing and Information Science Rochester Institute of Technology Rochester, NY USA Rochester Institute of Technology Service.
Virtualization Supplemental Material beyond the textbook.
Feifei Chen Swinburne University of Technology Melbourne, Australia
GreenCloud: A Packet-level Simulator of Energy-aware Cloud Computing Data Centers Dzmitry Kliazovich ERCIM Fellow University of Luxembourg Apr 16, 2010.
Accounting for Load Variation in Energy-Efficient Data Centers
Ensieea Rizwani An energy-efficient management mechanism for large-scale server clusters By: Zhenghua Xue, Dong, Ma, Fan, Mei 1.
Thin Clienting Justin Spratt. What is thin clienting? Thin clienting is a form of cloud computing—running applications on a server rather than on a local.
Tackling I/O Issues 1 David Race 16 March 2010.
Jennifer Rexford Fall 2010 (TTh 1:30-2:50 in COS 302) COS 561: Advanced Computer Networks Energy.
Low Carbon Virtual Private Clouds Fereydoun Farrahi Moghaddam, Mohamed Cheriet, Kim Khoa Nguyen Synchromedia Laboratory Ecole de technologie superieure,
Sql Server Architecture for World Domination Tristan Wilson.
Power Provisioning for a Warehouse-Size Computer (ISCA 2007) Authors: Xiabo Fan, Wolf-Dietrich Weber, and Luis Andre Barroso Google Presenter: Kirk Pruhs.
IIS Progress Report 2016/01/11. Goal Propose an energy-efficient scheduler that minimize the power consumption while providing sufficient computing resources.
GreenCloud: A Packet-level Simulator of Energy-aware Cloud Computing Data Centers Dzmitry Kliazovich, Pascal Bouvry, Yury Audzevich, and Samee Ullah Khan.
Energy Aware Network Operations
Lecture 2: Performance Evaluation
Virtual Machines What is a virtual machine?
Progress Report 2014/05/23.
Exploiting Sharing for Data Center Consolidation
4 steps to become green Patrick Pulvermueller
Green cloud computing 2 Cs 595 Lecture 15.
Inc. 32 nm fabrication process and Intel SpeedStep.
Multi-core CPU Power Control
ElasticTree Michael Fruchtman.
Cellular and Wireless Networks Power Management and Consumption
CHECO Fall 2013 Conference ERP Storage – What Works, What Does Not Presenter: Rick Beck, Director, IT Application Services September 17, 2013.
Virtualization.
The Greening of IT November 1, 2007.
The University of Adelaide, School of Computer Science
A workload-aware energy model for VM migration
IIS Progress Report 2016/01/18.
Presentation transcript:

Are Low Power Server CPUs Worth the Cost? Michael Fruchtman

Power Usage of a Small Server Setup Example: Given 4 servers Server consumes 250W each at 60% United States 10-11 amps, 1200W (safety margin) Cost $300 a year Europe 5amps (higher voltage), but amps cost more $500 a year Also - carbon footprint

In this corner…

The Judges vApus Mark I Simulates a consolidated virtualization load Meant for 90-100% load comparisons Good for HPC, not good for servers Servers operate at 30-60% Solution: Reduce the simulated machines by half Simulate load of an OLAP database (4 vCPUs), OLTP database (4 vCPUs), and 2 VMs Both CPUs can handle double this load Measure response time, throughput, and power usage

Idle Power Usage At idle PCU shuts down 5 cores Reduce clock speed No difference at idle

Under Load At average power use Throughput X5670 considerably more 25-60W more Throughput L5640 took 66 minutes X5670 took 59 minutes

Total Power Use Even power usage Race to idle L5640 and X5670 on balanced X5670 consumed 20Wh more Race to idle X5670 able to race to idle and shutdown Idle power consumption makes up for higher use

Throughput and Response Time Both graphs use the geometric mean.

Addition Statistics L5640 at Balanced to High Perf X5670 69% of throughput,76% longer response time, but 91% of power consumption L5640 at Balanced to Balanced X5670 76% of throughput, 34% longer response time, but 97% of power consumption Not a good deal if you can afford it X5670 costs $900 more than L5640 Adds only $200 a year in power cost

Any Advantages? Caps maximum power draw Range of X5670 is higher With L5640 one knows maximum draw of a server over time due to cap of 60W Data centers have penalties if one exceeds the power allowance Low power CPUs save power, but not energy

Questions