1 Automated Power Management Through Virtualization Anne Holler, VMware Anil Kapur, VMware.

Slides:



Advertisements
Similar presentations
Live migration of Virtual Machines Nour Stefan, SCPD.
Advertisements

Capacity Planning in a Virtual Environment
Overcoming the challenge of virtual blindness Colin Richardson on365 Ltd.
Office of Technology Operations & Planning Unlocking the Power of Server Virtualization Rebecca Astin Office of Technology Operations and Planning National.
Managing the Capacity and Performance of a VMware Cluster environment Presented by: Pete Weilnau CTO PERFMAN
VMware Virtualization Last Update Copyright Kenneth M. Chipps Ph.D.
Virtualization and Cloud Computing Virtualization David Bednárek, Jakub Yaghob, Filip Zavoral.
Adam Duffy Edina Public Schools.  The heart of virtualization is the “virtual machine” (VM), a tightly isolated software container with an operating.
Virtualization on the Intel Platform Scott Elliott Senior Systems Network Specialist Christie Digital A Customer Implementation with VMware and IBM.
Microsoft Virtual Server 2005 Product Overview Mikael Nyström – TrueSec AB MVP Windows Server – Setup/Deployment Mikael Nyström – TrueSec AB MVP Windows.
VIRTUALIZATION AND YOUR BUSINESS November 18, 2010 | Worksighted.
Commonwealth of Massachusetts Statewide Strategic IT Consolidation (ITC) Initiative ITD Virtualization and Shared Services Executive Briefing Presentation.
Virtualization for Cloud Computing
Virtualization Infrastructure Administration Cluster Jakub Yaghob.
Copyright © 2005 VMware, Inc. All rights reserved. VMware Virtualization Phil Anthony Virtual Systems Engineer
5205 – IT Service Delivery and Support
ProjectWise Virtualization Kevin Boland. What is Virtualization? Virtualization is a technique for deploying technologies. Virtualization creates a level.
Scalability Module 6.
Virtualization Performance H. Reza Taheri Senior Staff Eng. VMware.
VMware vSphere 4 Introduction. Agenda VMware vSphere Virtualization Technology vMotion Storage vMotion Snapshot High Availability DRS Resource Pools Monitoring.
Presented by : Ran Koretzki. Basic Introduction What are VM’s ? What is migration ? What is Live migration ?
PowerVM and VMware. What this presentation is Basic Terms that can be used to discuss multiple forms of virtualization Concepts common to virtualization.
WHAT IS PRIVATE CLOUD? Michał Jędrzejczak Główny Architekt Rozwiązań Infrastruktury IT
VMware Infrastructure 3 Overview. Agenda Introductions VMware Infrastructure 3 Overview Solutions, Use Cases Demonstration.
How to Resolve Bottlenecks and Optimize your Virtual Environment Chris Chesley, Sr. Systems Engineer
SAIGONTECH COPPERATIVE EDUCATION NETWORKING Spring 2010 Seminar #1 VIRTUALIZATION EVERYWHERE.
SAIGONTECH COPPERATIVE EDUCATION NETWORKING Spring 2009 Seminar #1 VIRTUALIZATION EVERYWHERE.
A Cloud is a type of parallel and distributed system consisting of a collection of inter- connected and virtualized computers that are dynamically provisioned.
Virtualization. Virtualization  In computing, virtualization is a broad term that refers to the abstraction of computer resources  It is "a technique.
Appendix B Planning a Virtualization Strategy for Exchange Server 2010.
Cloud Computing Energy efficient cloud computing Keke Chen.
VMware Infrastructure 3 The Next Generation in Virtualization.
Storage Management in Virtualized Cloud Environments Sankaran Sivathanu, Ling Liu, Mei Yiduo and Xing Pu Student Workshop on Frontiers of Cloud Computing,
Virtualization By Tim Ausburn & James Cantrell. Virtualization: Why? Reduce IT Costs Server consolidation Application Isolation Increase Server Utilization.
High Performance Computing on Virtualized Environments Ganesh Thiagarajan Fall 2014 Instructor: Yuzhe(Richard) Tang Syracuse University.
Never Down? A strategy for Sakai high availability Rob Lowden Director, System Infrastructure 12 June 2007.
Server Virtualization
Server Virtualization & Disaster Recovery Ryerson University, Computer & Communication Services (CCS), Technical Support Group Eran Frank Manager, Technical.
© 2012 IBM Corporation Platform Computing 1 IBM Platform Cluster Manager Data Center Operating System April 2013.
VGreen: A System for Energy Efficient Manager in Virtualized Environments G. Dhiman, G Marchetti, T Rosing ISLPED 2009.
Server Virtualization: Planning And Rollout George Crump, Founder Storage Switzerland.
VMware vSphere Configuration and Management v6
Copyright © 2005 VMware, Inc. All rights reserved. How virtualization can enable your business Richard Allen, IBM Alliance, VMware
Virtualization A brief introduction Virtualization1.
Cloud Computing Lecture 5-6 Muhammad Ahmad Jan.
Capacity Planning in a Virtual Environment Chris Chesley, Sr. Systems Engineer
Module Objectives At the end of the module, you will be able to:
© 2010 VMware Inc. All rights reserved Distributed Resource Management for Virtualized System Clusters VMware, Inc.
Practical IT Research that Drives Measurable Results 1Info-Tech Research Group Get Moving with Server Virtualization.
Unit 2 VIRTUALISATION. Unit 2 - Syllabus Basics of Virtualization Types of Virtualization Implementation Levels of Virtualization Virtualization Structures.
A Measured Approach to Virtualization Don Mendonsa Lawrence Livermore National Laboratory NLIT 2008 by LLNL-PRES
Unix Server Consolidation
Virtualization for Cloud Computing
A move towards Greener Planet
Bentley Systems, Incorporated
Workload Distribution Architecture
By Chris immanuel, Heym Kumar, Sai janani, Susmitha
Virtualization OVERVIEW
Virtualization Meetup Discussion
دکتر محمد کاظم اکبری مرتضی سرگلزایی جوان
Zhen Xiao, Qi Chen, and Haipeng Luo May 2013
Effective VM Sizing in Virtualized Data Centers
Virtualization.
Cloud Computing Architecture
Virtual Memory: Working Sets
Presentation transcript:

1 Automated Power Management Through Virtualization Anne Holler, VMware Anil Kapur, VMware

2 Agenda 1 Virtualization and Cluster Basics 2 Distributed Resource Management 3 Future Directions: Integrating Power Budget and Resource Management Across Virtualized Server Cluster

3 Agenda 1 Virtualization and Cluster Basics 2 Distributed Resource Management 3 Future Directions: Integrating Power Budget and Resource Management Across Virtualized Server Cluster

4 What is Virtualization?  Virtualization is a proven software technology that is rapidly transforming the IT landscape and fundamentally changing the way that people compute  Virtualization makes it possible to run multiple operating systems and multiple applications on the same SERVER at the same time Benefit  Virtualization allows you to reduce IT costs while increasing the efficiency, utilization and flexibility of your existing x86 computer hardware

5 For the visually inclined… Hypervisor

6 4 Key Properties  Partitioning  Isolation  Encapsulation  Hardware Independence

7 Partitioning Key Virtualization Properties  Run multiple operating systems on one physical machine  Divide system resources between virtual machines Hypervisor

8 Isolation Key Virtualization Properties  Fault and security isolation at the hardware level  Advanced resource controls preserve performance Hypervisor

9 Encapsulation Key Virtualization Properties  Entire state of the virtual machine can be saved to files  Move and copy virtual machines as easily as moving and copying files Hypervisor

10 Hardware Independence Key Virtualization Properties  Provision or migrate any virtual machine to any similar or different physical server Hypervisor

11 Live Machine Migration with Zero Downtime Enables the live migration of virtual machines from one host to another with continuous service availability. Benefits:  Revolutionary technology that is the basis for automated virtual machine movement  Meets service level and performance goals Hypervisor

12 Agenda 1 Virtualization and Cluster Basics 2 Distributed Resource Management 3 Future Directions: Integrating Power Budget and Resource Management Across Virtualized Server Cluster

13 DRS Overview Manage a Cluster As a Single Large Host 6 hosts CPU = 10 GHz Memory = 64 GB 1 “Giant Host” CPU = 60 GHz Memory = 384 GB

14 DRS Goals  Ease of Host Management  Initial Placement  CPU and Memory Load Balancing  VM-VM Affinity (Anti-Affinity)  VM-Host Affinity (Anti-Affinity)  Host Maintenance Mode Host Cluster vMotion Affinity

15 DRS Load-balancing Measuring Imbalance  Imbalance metric is a cluster-level metric  First for each host, DRS computes VM entitlement is a measure of how much resources a VM deserves (both CPU and memory). A host with higher capacity can have more VMs for the same normalized entitlement.  Imbalance metric is the standard deviation (spread) of these normalized entitlements Imbalance metric = 0.30 Imbalance metric =

16 Distributed Power Management Applying DRS to Power Management  Right-sizes Cluster Capacity  Consolidates virtual machines (VMs) onto fewer hosts & powers hosts off when demand is low  Powers hosts back on when needed to meet workload demand or to satisfy constraints  Optional add-on to Distributed Resource Scheduler DRS Cluster with DPM enabled Response to Reduced Demand Power Off

17 Agenda 1 Virtualization and Cluster Basics 2 Distributed Resource Management 3 Future Directions: Integrating Power Budget and Resource Management Across Virtualized Server Cluster

18 Motivating Problem  Rack power budget cannot be overcommitted or its fuse will trip  Equipment name plate max power draw is much larger than actually required (as much as 40%)  This leads to stranded (paid for) power and underutilizes available space in data center  Untapped power (and space) forces unnecessary DC purchases

19 Current (Partial) Solution  Vendors (e.g. HP, Dell, IBM) support server power capping to allow for safe over provisioning of rack allocated power  Determining appropriate power cap is a static, time consuming process to measure consumed power at each host and set cap  Vendor host power capping techniques do not take into consideration VM workload, which can clip performance so power cap today may impact tomorrow’s performance

20 Cloud Power Cap Solution: Determine Workload Needs Leveraging DRS  Load balance power across hosts in a rack to meet VM constraints and performance demands  Uses BMC setting which has host response rates in < 1 ms

21 Rack Power Budget Resource Trade-off CONFIDENTIAL Consider host comprising 12 CPUs, each 2.9 GHz, as follows: Power Cap(W) Num Hosts CPU Total (GHz) CPU RatioMemory Total (GB) Memory Ratio CPUMemoryNameplatePeakIdle 34.8 GHz96 GB400 W320 W160 W R ack w/8 KW power budget can support different maximum capacities CloudPowerCap manages racks w/host power cap below peak Trade-offs possible in allocating a rack’s 8KW power budget to its hosts:

22 Cloud Power Cap Distribution CONFIDENTIAL 22 VM 1 VM 2 cc VM 3 cc VM 3 VM 2 VM 1 cc Host A Host B Host A Host B Powercap Allocation Provide resources to allow CRM to satisfy VM-VM affinity constraint

23 Cloud Power Cap Entitlement Balancing Powercap Entitlement Balancing Enable entitlement balancing to enhance response to demand bursts (below) Reduce entitlement balancing overhead by avoiding some migrations VM 1 VM 2 cc VM 3 cc Host A Host B Host A Host B VM 1 VM 2 cc VM 3 cc

24 Cloud Power Cap Distribution Examples CONFIDENTIAL VM 1 Host A Host B Host A Host B VM 1 VM 2 cc Stand-by VM 2 CC Powercap Redistribution Provide burst headroom after host power-off during low demand period

25 Cloud Power Cap Design & Implementation Interacts with each CRM phase to support its goals  Operates on host/VM model used by CRM to determine actions  Modifies hosts’ CPU capacity when changing hosts’ power-cap  Issues actions in appropriate order wrt other CRM actions  Implemented w/Distributed Resource Scheduler (DRS)

26 Experiments Cloud Power Cap vs. Rack Budget Table Static High: Each host power cap statically set to its peak of 320W  Maximizes the amount of CPU capacity available for given rack power budget  Doesn’t enable using power for more memory & I/O capacity w/lower CPU Static: Each host power cap statically set to 250W  Allows more servers to be placed in the rack  Allows more memory and I/O capacity w/lower CPU capacity

27 Experiments CloudPowerCap vs. Static Strategies  Headroom balancing  Supporting DRS headroom balancing while avoiding vMotion overhead  Flexible resource capacity  Enabling trade-off between powering CPU and memory capacity at runtime

28 Headroom Rebalancing Experiment

29 Headroom Rebalancing Experiment

30 Flexible Resource Capacity Experiment

31 Flexible Resource Capacity Experiment

32 Questions & Answers