Delivering the next-generation data center to optimize business outcomes Date: Sep. 19 (Wed.) 김한결 과장 HP 2007 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
Challenge of Datacenter Techno-centric centric Business view of IT... a problem IT as a COST center Service-centric centric... a solution Shared Services 모델을 이용하여 IT 가 Business로 써 역할 Business- centric... a partnership IT as an BUSINESS innovation center Profit Center IT 조직은 기업에서 Business 조직과 같이 역할하며, Shared services의 창출, 판매, 중계, 제공을 통하여 높은 IT 투자 성과를 제공하는데 중점을 둠
Next Generation Datacenter 연중무휴 라이츠아웃(lights-out) 컴퓨팅 환경은 표준 시스템 확충방 식 (building block), 자동화된 맞춤 형(Modular) S/W, IT서비스를 위한 Supply Chain 구성에 달여 있다 낮은 IT 운영 비용 (Cost) 높은 서비스 수준 (Quality) 변화에 대한 빠른 대응 (Speed)
Shared Service Infrastructure Siloed Service Oriented Architecture of Shared Services Customers Employees Suppliers Managers Auditors Application Application Server Database Web Server Server Storage ROI Consolidate Refresh Right Size Virtualize Simplify Rationalize Standardize Modernize Modularize Shared Service Center Business Services Application Shared Services Application Infrastructure Shared Services Infrastructure Shared Service Utilities Technology Stack Business processes and projects Information Stack ~ Today (서비스 단위의 시스템 배치) ü 과거에는 시스템의 도입 시 하나의 서비스에 하 나의 시스템 형태로 도입 관리 됨에 따라, 관리의 효율성 측면을 중심으로 작업자의 작업 효율 및 동선에 고려하여 시스템을 배치 ü 주요 시스템 배치 기준 서비스별 배치 (작업자 중심의) 플랫폼을 기준으로 한 배치 Today ~ Future (데이터센터 관리 중심의 시스템 배치) ü 서비스의 빠른 변화에 적응하여 안정적으로 지원할 수 있는 IT아키텍처의 필요성은 가상화, 파티셔닝 등 Utility Computing 기술을 통한 시스템 통합 가속화 시키고 있음 ü 이에 따라 서비스와 Hardware와의 연관성이 느슨 해지고 과거와 같은 서비스단위의 상면 배치가 불가 능 ü 최근에 이러한 IT의존도 증가 및 시스템 통합으로 데이터센터의 관리 품질과 안정성이 중요하게 대두
Freeing IT investment for innovation 2002 2005 28% 23% 5% 45% 55% Application innovation ROI for Business & IT 72% 30% 42% 10% 15% 30% 45% Infrastructure innovation Application maintenance Infrastructure maintenance IT budget as % of revenue Source: HP-IT
Consolidate Datacenter How to make cost reduction? Adaptive Infrastructure
How to make the Profit Center - Shared Service 경영 프로세스 핵심역량 감량 경영 Shared Service는 기업의 기능 중 일부를 모기업과 별도의 경영구조를 갖는 반 자치사업 단위로 집중시켜 내부고객인 모기업의 능률향상, 가치 창출, 비용절감 및 서비스 개선을 도모하기 위한 일종의 협업 전략 프로세스 모기업 Shared Service 단위 모기업에 속한 부분 단위이기보다는 공개시장에서 경쟁적으로 이루어지는 비즈니스 형태와 유사할 수 있음. Bryan P. Bergeron Essentials of Shared Services Shared Infrastructure 의 Value chain 흐름 Requests for access Demand application Services receive orders Send deliveries Issue Work order Planning and admin Operation Report receipt delivery Send an order Receive delivery Requests for storage Supply Storage SLA Charge back Utility Computing Value decoupling
How to make the Profit Center - Consolidated Datacenter 전산센터는 기업내의 비즈니스를 기준으로 한 분산형 데이터센터 구조에서 Shared Service 모델에 적합한 중앙 집중적 데이터센터 체계로 변화 해나가고 있습니다. AS-IS 분산형 데이터센터 개별 비지니스 중심의 지역별 전산센터 TO-BE 비지니스 공용의 전산센터 중앙집중적 데이터센터 Mega Datacenters Corporate 1 Datacenter 2 Corporate 1 Corporate 1 Datacenter 1 Datacenter 3 Corporate 1 Corporate 2 Corporate 3 Corporate 4 *)Business Continuity Datacenter 5 Corporate 1 Datacenter 4 Corporate 1 Region Center Region Center Region Center Machine room Machine room Machine room
제 2기 Global Datacenters - Hewlett Packard IT Challenge Solution Target State 너무 많은 프로젝트가 동시에 진행됨에 따라 수행 속도의 향 상이 요구됨 IT 자원이 50 여개국 100개의 사이트 배치가 되어서 통합이 요구가 됨 기업 인원 및 리소스가 기본 활 동에 너무 편중되어 기업 혁신 프로젝트의 진행 의 어려움 발 생 기업 비즈니스 정보가 100 여 개가 넘는 데이터 마트에 산재 통합 필요 IT 환경의 거대화 및 복잡도 증 가에 따른 비용 증가 다섯가지의 전략을 기반으로 변화를 수행하였습니다. 사업의 포트폴리오 효율 화 IT업무 환경 효율화 글로벌 데이터센터 전사적 데이터 웨어하우 스 World-class IT Simplicity 비즈니스 중요도를 기반으로 한 IT 프로젝트 중요도 재정립 핵심 비즈니스 어플리케이션 으로 표준화 및 불필요 요소 제거 Efficiency 중앙집중적 IT 조직 및 비용구조 마 련 (Shared Service) IT 인력에 대한 Core Site로 집중화를 통한 업무 향상성 증대 Agility 업무 역량의 재할당: 80% 혁신, 20% 유지 통합 엔터프라이즈 DW 구축 Value 비용 절감: 6개의 글로벌 데이터센터 구축 연간 1조의 운영비용 절감 대외 홍보 위한 HP 제품 및 전산센터 통합 서비스 모델 마련
HP Global Datacenters Zone1 Zone3 Site 1 Austin Atlanta Site 5 Site 2 Houston Site 6 Site 3 Site 4 Zone2 ü 3 Zones supporting HP s Global Applications (Austin, Houston, Atlanta) ü Total white space ~410,000 sq ft (서버 룸 공간 기준 약 1,1000평) ü Within each zone: 2 sites within 10-25 mile radius of each other where feasible ü Each site designed for High Availability, Disaster Recovery and Business Continuity
HP DCC Considerations & Scope Consolidated Infrastructure Secure facilities Wide area Network Redesign of DRP & BCP Shared infrastructure Site Location Building & Physical Environment Consolidated service helpdesk How many Datacenter Moving Equipment Physically consolidated datacenter Remained equipment & resource Down time Investment plan Expected budget Expected ROI (Return on Investment) ROI Measurement system IT governance model Ensure business alignment Proper balance of cost, service levels, risk and agility for the business Must support standards and shared infrastructure IT processes Many processes will need to change because of shared infrastructure, and to maintain the efficiencies from consolidation IT organization Reorganization may be warranted given other changes, especially with collocation HR Planning (retention, reduction etc) Human impact Consolidation means change for your staff May mean role change, retooling, relocation or redeployment Retention and buy-in may be factors
Space Design - Modular Datacenter Concept Zone Building Module Pod Facility Availability : Tier 3 +
Business Continuity Global Datacenters Business Continuity Global Data Centers Next generation data centers Tier 3+ Site 1 Site 2 Zone-1 Application Zone-2 Next generation data centers Tier 3+ Site 3 Site 4 Active Dark Active Dark Tier 1 Active Dark Active Dark Active Dark Active Dark Tier 2 Active Dark Active Dark Active Dark Active Dark Tier 3 Active Dark Active Dark Business Disaster Business Availability continuity Reliability recovery Expandability continuity Agility
Strategy - HP Datacenter Lifecycle Service Implement of Virtualization & Automation Infra Enhancement Service Oriented Architecture & Standardization of IT architecture Service Enhancement Next Gen Datacenter Build up governance Infrastructure Management Standardization Governance Data center Lifecycle Consolidate / Relocate datacenter. Re-align IT Resource. Datacenter Transition Datacenter Build Datacenter Master Plan Strategic Planning Build up Datacenter & Environment Infrastructure Design blue print for Next Generation Datacenter Align IT Investment Plan with Biz Requirement
Service Offering - Datacenter Service HP Data Center Services 어떻게 하면 적기에 위험을 최소화하면서, 최상의 프로젝트를 제공할 것인가? 어떻게 하면 비용을 절감하면서 최적의 서비스를 받을 수 있을 것인가? 어떻게 하면 데이터센터의 기반환경부터 서비스 까지 전반적인 성능 향상을 가져갈 것인가? Project Services Operation Services Optimization Service Migration Service Site Planning Service Relocation Service Deployment Service Project Management Service üconceptual Engineering Service üinfrastructure Construction Service ühp Solution Implementation Service Server Operation Service Storage Operation Service Software Operation Service Network Operation Service Resource and Technical assistance ITSM assessment & design service IT Security and Risk assessment & design service Datacenter assessment & design service Power & Cooling Service Infrastructure optimization assessment & implementation service
Datacenter Infrastructure How to make more energy efficiency? Adaptive Infrastructure Power & Cooling Key Enabler
Datacenter Facility Key Factors Trade-off Power Power/ cooling interaction Cooling Trade-off Power density Datacenter layout Trade-off Space
IT System power densities Trend System Product Lifecycle : 3~5 years Datacenter Product Lifecycle : 10 ~ 25 years 신규 데이터센터 구축 시 시스템의 매래 추이를 고려하여 설계 되어야 함. Source: Datacom Equipment Power Trends and Cooling Applications, ASHRAE, 2005, http://www.ashrae.org Ver 1.0 (2000) TUI KWatt/m 2 New Version (2004) KWatt/m 2 2000 2010 ASHARE 2000 2010 Server Disk Storage System 10 23 Compute Servers 1U Blade and Custom Computer Server 2U and Greater 15.4 64.9 35.2 Storage 7.7 19.8 Source : Data Equipment Power Trend Charts : TUI, 2000 Communication Equipment 20 100 Communication High Density Communication Extreme Density 15.4 42.9 26.4 110
Required Holistic Management Holistic View from the Chip Core to the Cooling Tower W b W b W pump W compressor W pump Q system W sys data center is the computer IT + Facility View Q chip W chip Total Heat Dissipation COP = = Heat Extracted by Air Conditioners Net Work ( Flow Work + Thermodynamic Work) of Cooling system Input Patel et. al, Energy Flow in the Information Technology Stack: Coefficient of Performance of the Ensemble and its Impact on TCO http://www.hpl.hp.com/techreports/2006/hpl-2006-55.html
Low Efficiency in Cooling Environment 38 C Vertical Under Floor 100 23 C Rack Heat Loads 8KW 43 C 43 C Rack Heat Loads 8KW 100 23 C Ideal Case 정리되지 않은 케이블 블랭크 판넬 미사용 23 C 23 C CFM 이중마루 임의 절단 36 C Hot Spot 100 공급 공기 온도 23 C Front Rack Heat Loads 8KW 배출 공기 온도 43 C Rear F-R-F-R 형태의 잘못된 랙 배치 100 23 C Rack Heat Loads 8KW Rack Heat Loads 43 C 36 C 58 C 8KW 상기 조건은 시스템의 CFM / 발열량에 따라 변화할 수 있습니다. Hot Spot 40 Rack Heat Loads 23 C Rack Heat Loads Hot Spot 40 이중마루 하부의 케이블 적체 22 C 8KW 8KW 22 C
Best Practice Cooling Environment VUF Based Extended Cooling Solution Supplemental Floor Mount 0 5 10 15 20 25 30 35 KW/ Rack Floor Mount Vertical Under Floor Best Practice Supplemental VUF with Supplemental Extended Cooling Solution Chiller with High-density Cooling
HP Datacenter Strategy Sustainable computing is the goal Industrialization of IT Data centers as building blocks It s about profit per watt Data center energy will be a public policy issue Integrated Infrastructure Convergence & automation Value migration: HW to SW Control engines & tools (policies) Global workforce, global IT Sustainability Energy efficiency & de-materialization Industry leads or government regulates Compute Power Cooling Control Engines, Tools Sensing infrastructure Flexible & configurable The data center is the computer
HP Power and Cooling research History in the data center 1996 Computational Fluid Dynamics (CFD) applied to internal data centers 2001 Led the drafting of the Thermal Guidelines for Data Processing Environments and is one of the primary contributors. 2003 Smart Cooling Thermal Assessments introduced 2006 C-Class Blades introducing HP Thermal Logic for improved efficiency 2007 Thermal Zone Mapping introduced 1999-2000 HP embraced the Uptime Institute s Fault Tolerant Power Compliance Specification for servers. 2002 First to introduce data center metrics which analyzed the corelation between racklevel recirculation of hot air and the shortcircuiting of cool air at air-conditioner inlets. 2004 Introduction of Sever power management technologies such as power regulator 2006 Modular Cooling System (MCS) Liquid cooled Rack 2006 Dynamic Smart Cooling announced
HP driving industry efficiency metrics - PUE Broad industry adoption Server efficiency - April 2006 1 Industry used this work as foundation for: Energy Star Paper (HP a primary) 2 SPEC Power Benchmark (HP Chairs Committee) Data center efficiency April 2006 3 Power Usage Effectiveness ratio (PUE) has been adopted by: ASHRAE (HP is a committee Founder) Green Grid (HP is member of the BoD and Metrics Work Group Chair) Power Consumption 100% Average Perf/Watt at each point 0% 20% 40% 60% 80% 100% Performance PUE = Total Facility Power / IT equipment power Building load Demand from grid Power (Switch Gear, UPS, Battery backup, etc) Cooling (Chillers, CRACs, etc) IT load Demand from servers, storage, telco equipment, etc Sources: 1 http://thegreengrid.org/pdf/efficiency_slides_for_general_distribution_final.pdf 2 http://www.energystar.gov/ia/products/downloads/finalserverenergyprotocol-v1.pdf 3 C. Malone, C. Belady, "Metrics to characterize Data Center & IT Equipment Energy Use, Proceedings of 2006 Digital Power Forum, Richardson, TX (September 2006)
HP leadership through holistic approach Holistic offering (chips to data center) Efficient systems, flexible designs Leadership in innovation Chips Systems and software Infrastructure Data center Low voltage Multi-core HP Bladesystem design Cluster platform configurations Server and storage consolidation SIM and Openview Universal racks, PDUs, UPSs Modular Cooling System Mission critical services Data center services Site planning/relocation Thermal assessment Dynamic Smart Cooling OS features Virtualization
Smart Data Center (SDC) Technology Roadmap Dynamic Smart Cooling (DSC) = data center cooling management SDC = DSC + Power Regulator (IT equipment power management) Static & Modular Static Data Center Optimization Static Smart Cooling Modular Cooling System (MCS) Dynamic Managed Data Center Cooling Dynamic Smart Cooling Low-cost temperature sensor network Next-gen MCS Rule-Based Smart Data Center V1.0: Synchronized Management of P&C Power Regulator for Integrity and ProLiant DC-level P&C management with SIM & gwlm (Power Regulator + Dynamic Smart Cooling) Rack-level power capping via SIM Automated at App Smart Data Center V2.0: P&C Management for NGDC SLA-based DC P&C management with OpenView DC-level P&C management for liquid-cooled servers and racks P&C-driven workload migration RF-ID-enabled visualization tools Power-aware software applications Today 2007 2008 2009 +
Service Offering - Datacenter Infrastructure Service HP Data Center Infrastructure Services Thermal Assessment Datacenter Assessment Datacenter Design & Plan Service 전산센터 내의 공조 환경 분석 및 개선 작업 수행 Temperature & Airflow Simulation (CFD) Infrastructure standardization Cooling infra Renovation 전산센터 전반의 상태 점검 및 개선 작업 수행 Facility, Operation assessment Certification Renovation 전산센터 구축을 위한 사전 개념 설계 작업 수행 Capacity planning Infrastructure Architecture design Time Schedule Datacenter Build Service Relocation Service Project Management Service 전산센터의 인프라 구축을 위한 Multi-vendor contract Integrated support Project management HP 장비 뿐만이 아닌 전산실 내부의 모든 시스템에 대하여 Plan (On-line, Off-line) Implementation 데이터센터와 관련이 있는 사업에 대하여 Plan Control Management Multi-vendor contract
Sample - Datacenter Thermal Assessment (3D Cooling) Cooling / Airflow Assessment Part 상태 : Hot Spot Zone 발생 향후 예상 되는 장애 요소 : CRAC 증설 및 Blank Panel 추가 Airflow Design 방안 제시
Case Study: HP Palo Alto, Global Datacenters Before assessment CRAC 2, 100% @ 18 C CRAC 3, 75% @ 19 C CRAC 4, 100% @ 13 C CRAC 5 off CRAC 6 off CRAC 1, 100% @ 18 C Assumed Best Practices we are as efficient as we can be Savings (Extrapolated) 20,000 SF data center $500K per year 4.5 million lbs CO2 emissions equivalent: about 900 round trips NY to S Francisco in an SUV This is a straight line extrapolation for illustration purposes only to show the scale of what may be achieved DC Optimized using Thermal Assessment data CRAC 2, 50% @ 21 C CRAC 3, 100% @ 21 C CRAC 4, 50% @ 21 C CRAC 5 off CRAC 6 off CRAC 1, 100% @ 21 C 25% reduction in cooling consumption
Q & A