Power Management System by sej73010

VIEWS: 6 PAGES: 24

More Info
									Dynamic Power Management
in the Data Center
IBM Power and Cooling Symposium
Raleigh, NC
04 October 2007
Bharat S. Shah
Emerson Network Power
Bharat.Shah@EmersonNetworkPower.com
Agenda
 Power Management Challenges
 IT Equipment Power Reduction Schemes
 Facility Infrastructure Power Improvements
 Dynamic Power Management System
 Summary




    04 Oct 2007                       Slide 2
Power Management Challenges
 Over the next five years, power
 failures and limits on power
 availability will halt data center
 operations at more than 90% of
 all companies (AFCOM (DCI))
 By 2008, 50% of data centers
 will have insufficient power and
 cooling capacity to meet the
 demands of higher density
 equipment (Gartner)
 Power and cooling spending will
 exceed server spending within 2
 years (IDC, Gartner)
 Power associated with servers,
 storage and network equipment
 will increase from 1.5% of total     Source: EPA, Andrew Fanara

 US energy consumption to 2.5%
 by 2011 (EPA)


      04 Oct 2007                                                  Slide 3
Current Data Center Power Consumption


       Cooling
        25%
                                                            Servers
    Air                                                      42%
 Movem  ent
   12%
                                             ther
                                            O IT
            Electricity                    Equipm ent
                      er
           Transform Lighting,                8%
               UPS       etc.
               10%       3%

Source: EYP Mission Critical Facilities Inc., Emerson estimate

                                                                      Source: Green Grid
                                                                      Feb 2007




              04 Oct 2007                                                            Slide 4
 Current Data Center Power Management
                               No System Linkage
                               No System Linkage
                         -Workflow Processes at Best
                        -Workflow Processes at Best
                         -Typically E-Mail/Phone Communication
                        -Typically E-Mail/Phone Communication
                         -Reactive Mode at Worst
                        -Reactive Mode at Worst




         IT                                                    Facility
     Management                                              Management
       System                                                  System



Servers Routers 3rdParty                                 UPS Cooling 3rd Party
    Storage Switches                                        ATS GenSets



          04 Oct 2007                                            Slide 5
Agenda
 Power Management Challenges
 IT Equipment Power Reduction Schemes
 – Server Hardware
 – Virtualization Software
 – Dynamic Efficiency Control




    04 Oct 2007                   Slide 6
Server Hardware Architecture Example
                                                       Dynamically
                                                        optimize
                                  IBC
   AC-DC / PFC
                                                        efficiency
   Power Supply
                                        POL Reg
                                         DC-DC/VRM

                          6-12                                    Processor,
                     48      12            POL Reg
                                           DC-DC/VRM
                                                                   memory
                    Vdc     Vdc               POL Reg
                                               DC-DC/VRM
                                                                   and I/O
                                                                   loading
                                                 POL Reg
                                                  DC-DC/VRM


 Dynamically
 optimize IBV




      04 Oct 2007                                       Slide 7
 Virtualization Software




Source: NewsCorp
IDC Virtualization Forum, Feb 2006
              04 Oct 2007            Slide 8
  Dynamic Efficiency Control
  Digital Power Supplies
       Dynamic efficiency enhancement based on load
       fluctuation thru virtualization (via digital power supplies)
                                                    High Line (230V) Efficiency

                                95
                                93
                                91
                                89
                   Efficiency




                                87
                                                                                             Today
                                85
                                                                                             Optimized
                                83
                                81
                                79
                                77
                                75
                                     10   20   30   40    50     60     70   80   90   100
                                                         Percent Load


        Efficiency Curve Shaping: This is an optimized 760W 1+1 power supply for
        a data center application. It is optimized for efficiency at 20% to 50% load –
        the range in which it will operate for 98% of its service life.
Source: Emerson Internal


               04 Oct 2007                                                                      Slide 9
 Dynamic Efficiency Control
 PMBus™ Communication
 PMBus™

                                                    Data Center Applications


                                                      System Applications


         AC/DC PSU                      WSMAN             Motherboard
                              PMBusTM     Management
                                                                            Components
                                          Controller(s)
         AC/DC PSU




      PMBus™ compliant power supplies provide in-band /
      out-of-band access to sensors, reporting power
      consumption data via SMBus
      GUI displays system power consumption reported by
      Digital Power Supply


Source: Intel
Intel Developers Forum, Apr 2007
              04 Oct 2007                                                       Slide 10
Agenda
 Power Management Challenges
 IT Equipment Power Reduction Schemes
 Facility Infrastructure Power Improvements
 – Digitally-Controlled Compressors
 – Scalable UPS Systems
 – Plug-and-Play Cooling Capacity




    04 Oct 2007                       Slide 11
Digitally-Controlled Compressors
Digitally-Controlled
                                                             Average Compressor Cycles per Year
      Annual Compressor Cycles
                                 35000
                                 30000
                                 25000
                                 20000
                                 15000
                                 10000
                                  5000
                                    0

                                                                 Two compressor




                                                                                                                         Two compressor
                                                                                   Two compressor




                                                                                                     Single compressor
                                         Single compressor




                                                                                    with unloading




                                                                                                                          fully variable
                                                                   Dual circuit




                                                                                     Dual circuit




                                                                                                                           Dual circuit
                                                                                                        fully variable
                                            Single circuit




                                                                                                        Single circuit



                                                                                                                             (Digital)
                                                                                                           (Digital)
                                                                                         (SH)
                                                                                  System Type



       Continuously variable capacity compressors
  dramatically reduce the frequency of start/stop cycles –
     saving energy and significantly increasing MTBF.
     04 Oct 2007                                                                                                                   Slide 12
Scalable UPS Systems


 Capacity added via software key
 (enables additional capacity
 without hardware addition/change)
 Examples:
  – Soft-Scale (fine adjustment)
      • 40kVA → 60kVA → 80kVA
      • 80kVA → 100kVA → 120kVA
      • 120kVA → 160kVA → 200kVA


                                     80 kVA – 100 kVA – 120 kVA

      Scalable systems enable pay-as-you-go capacity
      without over-engineering for maximum capacity –
          saving equipment capex and energy opex.
      04 Oct 2007                                 Slide 13
Plug-and-Play Cooling Capacity
Plug-and-Play
                             Pre-installed Piping      Quick-Connect




                                                       Supplemental
                                                       Cooling

     Supplemental cooling can be added to hot-spots
        saving energy and minimizing labor costs.

     04 Oct 2007                                    Slide 14
Agenda
 Power Management Challenges
 IT Equipment Power Reduction Schemes
 Facility Infrastructure Power Improvements
 Dynamic Power Management System




    04 Oct 2007                      Slide 15
 Dynamic Power Management System
 Bridging/Optimizing IT and Facility Systems

                         Power Management System
                         Power Management System
                    -Centralized or Distributed Intelligence
                   -Centralized or Distributed Intelligence
                    -Optimization Policies Business Rule Engine
                   -Optimization Policies / /Business Rule Engine
                    -Manual or Auto-Pilot Operation
                   -Manual or Auto-Pilot Operation


                        Open Bus                   Open Bus



         IT                                                       Facility
     Management                                                 Management
       System                                                     System
                                                  Proprietary
                                                      or
                         Open                                                  Open
                                                    Open

Servers Routers 3rdParty                                   UPS Cooling 3rd Party
    Storage Switches                                          ATS GenSets



          04 Oct 2007                                               Slide 16
Dynamic Power Management System
 Reduce Energy Costs
  – Equipment power control (on/off/standby) and
    equipment provisioning in response to dynamic
    computing demand
 Reduce Cooling Needs
  – Balanced cooling based on dynamic IT workload
    conditions
 Improve Capacity Planning
  – Scenario planning of power/cooling infrastructure
    changes in response to computing load changes



     04 Oct 2007                            Slide 17
Dynamic Power Management Systems
Reduce Energy Costs


                                             Optimize energy
                    ATS/
                   Gen S
                         ets                 consumption across
                                 System      the entire data center
                       stems                 in response to IT
              U   PS Sy        Integration
                                             workload variations
                 g Sys
                      tems                   Capture additional
           Coolin
                                             savings based on
                        /
                 ervers pplies
                                             utility availability or
               S
                Power
                       Su                    rate structure
     D   igital


                    Optimize power consumption
                        from CHIP to GRID.

    04 Oct 2007                                         Slide 18
Dynamic Power Management Systems
Reduce Cooling Needs




 Set rules to adapt cooling to the load
  –   Manage compressor load across units
  –   Manage humidification & dehumidification
  –   Cycle units, rotate standby units


                     Further reduce energy costs and
                     extend life of cooling equipment.

       04 Oct 2007                                       Slide 19
Dynamic Power Management Systems
Improve Capacity Planning

 Match power / cooling
 capacity with computing
 resource consumption
  – Determine if additional
    computing resources can
    be added within a specific
    power and cooling budget
    in the same physical space
  – Or if new space is required
    to house these additional
    computing resources


    Avoid reactive crises through proactive planning of
     computing and infrastructure change scenarios.

     04 Oct 2007                                 Slide 20
Dynamically Optimized Data Center
                                                                                  2008,2009 - ----
                                                                                  •Integrated, dynamically optimized
                                                                                  adaptive systems
                                                                              t
               Integrated



                                                                           en
                                                                       ge
                                                                          m
                                                                                  •Virtualization and load balancing
                                                                   ana
                                                              rM                  linked to power and cooling
                                                            we
                                                          Po
                                                    ith
                                  2007-2008     W                                 •Intelligent models for control and
                                                                                  Optimization/Advance fault notification
Provisioning




                                   Real time monitoring
                                                                                                                           t
                                                                                                                        en
                                Now                                                                                 ge
                                                                                                                       m
                                                                                   2007-                          na
                                •Modular systems                                                             M
                                                                                                              a
                                                                                   2008                   er
                  Separate




                                •Adaptive scalable                                                  Po
                                                                                                      w

                                architecture                                                  ith
                                                                                  Sensing W
                                •Separate systems for                                &
                                power, cooling,…..                                Control


                                      Monitoring & Limited                                           Adaptive Intelligent
                                      Conventional Controls                                              Controls
                                                                       Type of Control




                             04 Oct 2007                                                                                       Slide 21
Agenda
 Power Management Challenges
 IT Equipment Power Reduction Schemes
 Facility Infrastructure Power Improvements
 Dynamic Power Management System
 Summary




    04 Oct 2007                      Slide 22
Summary
 IT and Facility equipment continue to evolve
 intelligently
 – Higher Efficiency, Digital Control and Virtualization
 Current management, monitoring and control
 software optimize IT and Facility systems
 independently
 Bridging the IT and Facility systems with an
 integrated Power Management solution
 dynamically optimizes efficiency across the entire
 Data Center


    04 Oct 2007                              Slide 23
Power Operating System
Conceptual Data Flow




     04 Oct 2007         Slide 24

								
To top