Lean Computing for the Cloud

ERIC BAUER is Reliability Engineering Manager in the IP Platforms Group of Alcatel-Lucent and Bell Labs Fellow. Before focusing on reliability engineering, Mr. Bauer spent two decades designing and developing embedded firmware, networked operating systems, internet platforms, and optical transmission systems. He has been awarded more than twenty US patents, and has authored several books such as (2013) Service Quality of Cloud-Based Applications, (2012) Reliability and Availability of Cloud Computing, and (2010) Design for Reliability: Information and Computer-Based Systems, all of which were published by Wiley-IEEE Press. Mr. Bauer earned his BS in Electrical Engineering from Cornell University and MS in Electrical Engineering from Purdue University.

• Introduction xi Acknowledgments xvAbbreviations xvii1. Basics 11.1 Cloud Computing Fundamentals 11.2 Roles in Cloud Computing 61.3 Applications 91.3.1 Application Service Quality 111.4 Demand, Supply, Capacity, and Fungibility 131.5 Demand Variability 161.6 Chapter Review 182. Rethinking Capacity Management 192.1 Capacity Management 192.2 Demand Management 212.3 Performance Management 212.4 Canonical Capacity Management 232.4.1 Traditional Capacity Management 242.4.2 ITIL Capacity Management 272.4.3 eTOM Capacity Management 282.4.4 Discussion 302.5 Three Cloud Capacity Management Problems 302.5.1 Physical Resource Capacity Management 312.5.2 Virtual Resource Capacity Management 322.5.3 Application Capacity Management 332.6 Cloud Capacity Management as a Value Chain 362.7 Chapter Review 393. Lean Thinking on Cloud Capacity Management 413.1 Lean Thinking Overview 413.2 Goal 423.3 Seeing Waste (Nonvalue-Adding Activities) 433.3.1 Reserve Capacity 453.3.2 Excess Application Capacity 463.3.3 Excess Online Infrastructure Capacity 463.3.4 Excess Physical Infrastructure Capacity 463.3.5 Inadequate Capacity 473.3.6 Infrastructure Overhead 483.3.7 Capacity Management Overhead 483.3.8 Resource Overhead 493.3.9 Power Management Overhead 503.3.10 Workload Migration 503.3.11 Complexity Overhead 513.3.12 Resource Allocation Failure 513.3.13 Leaking and Lost Resources 533.3.14 Waste Heat 533.3.15 Carbon Footprint 543.4 Key Principles 543.4.1 Move toward Flow 553.4.2 Pull versus Push 553.4.3 Level the Workload 553.4.4 Stop and Fix Problems 553.4.5 Master Practices 563.4.6 Visual Management 573.4.7 Use Well-Tested Technology 573.4.8 Take a Long-Term Perspective 583.4.9 Grow, Learn, and Teach Others 583.4.10 Develop Exceptional People 583.4.11 Partners Help Each Other Improve 583.4.12 Go See 593.4.13 Implement Rapidly 593.4.14 Become a Learning Organization 593.5 Pillar: Respect 593.6 Pillar: Continuous Improvement 613.7 Foundation 623.8 Cadence 623.9 Lean Capacity Management Philosophy 633.10 Chapter Review 644. Lean Cloud Capacity Management Strategy 674.1 Lean Application Service Provider Strategy 684.1.1 User Workload Placement 714.1.2 Application Performance Management 734.2 Lean Infrastructure Service Provider Strategies 734.2.1 Physical Resource Capacity Management 764.3 Full Stream Optimization 774.4 Chapter Review 795. Electric Power Generation as Cloud Infrastructure Analog 815.1 Power Generation as a Cloud Infrastructure Analog 815.2 Business Context 835.3 Business Structure 865.4 Technical Similarities 885.5 Impedance and Fungibility 915.6 Capacity Ratings 945.7 Bottled Capacity 955.8 Location of Production Considerations 955.9 Demand Management 975.10 Demand and Reserves 985.11 Service Curtailment 995.12 Balance and Grid Operations 1005.13 Chapter Review 1036. Application Capacity Management as an Inventory Management Problem 1056.1 The Application Capacity Management Service Delivery Chain 1056.2 Traditional Application Service Production Chain 1076.3 Elasticity and Demand-Driven Capacity Management 1086.4 Application Service as Retail Analog 1106.4.1 Locational Consideration 1126.4.2 Inventory and Capacity 1126.4.3 Service Level 1136.4.4 Inventory Carrying Costs 1146.4.5 Inventory Decision, Planning, and Ordering 1156.4.6 Agility 1186.4.7 Changing Consumption Patterns 1186.5 Chapter Review 1187. Lean Demand Management 1197.1 Infrastructure Demand Management Techniques 1207.1.1 Resource Scheduling 1217.1.2 Resource Curtailment 1217.1.3 Mandatory Demand Shaping 1227.1.4 Voluntary Demand Shaping 1237.1.5 Scheduling Maintenance Actions 1237.1.6 Resource Pricing 1237.2 Application Demand Management Techniques 1247.2.1 Queues and Buffers 1247.2.2 Load Balancers 1247.2.3 Overload Controls 1257.2.4 Explicit Demand Management Actions 1257.2.5 Scheduling Maintenance Actions 1257.2.6 User Pricing Strategies 1267.3 Full Stream Analysis Methodology 1267.3.1 Analyze Applications' Natural Demand Patterns 1277.3.2 Analyze Applications' Tolerances 1287.3.3 Create Attractive Infrastructure Pricing Models 1297.3.4 Deploy Optimal Infrastructure Demand Management Models 1307.4 Chapter Review 1318. Lean Reserves 1338.1 What Is Reserve Capacity? 1338.2 Uses of Reserve Capacity 1358.2.1 Random Demand Peaks 1358.2.2 Component or Resource Failure 1368.2.3 Infrastructure Element Failure 1368.2.4 Infrastructure Resource Curtailment or Demand Management Action 1378.2.5 Demand Exceeding Forecast 1378.2.6 Lead Time Demand 1378.2.7 Catastrophic Failures and Force Majeure Events 1398.3 Reserve Capacity as a Feature 1398.4 Types of Reserve Capacity 1408.4.1 Automatic Infrastructure Power Management Controls 1408.4.2 Utilize Application Reserve Capacity 1418.4.3 Place/Migrate Demand into Underutilized Capacity 1418.4.4 Grow Online Capacity 1418.4.5 Service Curtailment/Degradation 1418.4.6 Mandatory Demand Shaping 1418.4.7 Voluntary Demand Shaping 1428.4.8 Emergency Reserves 1428.5 Limits of Reserve Capacity 1448.6 Ideal Reserve 1448.6.1 Normal (Co-located) Reserve 1448.6.2 Emergency (Geographically Distributed) Reserve 1468.7 Chapter Review 1479. Lean Infrastructure Commitment 1499.1 Unit Commitment and Infrastructure Commitment 1509.2 Framing the Unit Commitment Problem 1519.3 Framing the Infrastructure Commitment Problem 1539.4 Understanding Element Startup Time 1559.5 Understanding Element Shutdown Time 1579.6 Pulling It All Together 1609.7 Chapter Review 16610. Lean Cloud Capacity Management Performance Indicators 16710.1 Perfect Capacity Metrics 16810.2 Capacity Management Metrics 17210.3 Infrastructure Commitment Metrics 17310.4 Waste Metrics 17410.4.1 Reserve Capacity Waste Metrics 17410.4.2 Excess Application Capacity Metrics 17510.4.3 Excess Online Infrastructure Capacity Metrics 17510.4.4 Excess Physical Infrastructure Capacity Metrics 17510.4.5 Inadequate Capacity Metrics 17510.4.6 Infrastructure Overhead Waste Metrics 17610.4.7 Capacity Management Overhead Waste Metrics 17610.4.8 Resource Overhead Waste Metrics 17610.4.9 Power Management Overhead Waste Metrics 17710.4.10 Workload Migration Metrics 17710.4.11 Complexity Overhead Metrics 17810.4.12 Resource Allocation Failure Metrics 17810.4.13 Leaking and Lost Resources 17910.4.14 Waste Heat Metrics 17910.4.15 Carbon Footprint Metrics 18010.5 Key Principle Indicators 18010.6 Cost of Poor Quality 18110.7 Metrics and Service Boundaries 18210.8 Measurements and Maturity 18310.9 Chapter Review 18511. Summary 18711.1 Cloud Computing as a Service Delivery Chain 18711.2 Lean Cloud Computing 19011.3 Reimagining Cloud Capacity 19211.4 Lean Demand Management 19511.5 Lean Reserves 19711.6 Lean Infrastructure Service Provider Considerations 19811.7 Lean Application Service Provider Considerations 19811.8 Lean Infrastructure Commitment 19911.9 Visualizing Perfect Capacity 20111.10 Lean Cloud Computing Metrics 20311.11 Concluding Remarks 204References 207About the Author 211Index 213