Cloudonomics: Strategy & Statistics

Joe Weinman
SVP, Cloud Services and Strategy, Telx jweinman@telx.com @joeweinman
2013 Telx. All Rights Reserved.

Cloudonomics

Questionable Notions
The cloud is a revolutionary new technology and business model Cloud services are computing resources delivered over the Internet The primary benefit is cost reduction; the other is business agility Economies of scale are the key to cloud benefits Another unique benefit is the ability to replace CapEx with OpEx Rational decision makers will select the lowest cost option Scale economies consolidation to a handful of providers Big Switch commoditization, all IT moving to the cloud IT Doesnt Matter cost optimization is key to provider selection Cloud reduces computing costs reduced IT spend However, data center electricity consumption is too high data centers dont create jobs and security is an issue
Source: Joe Weinman, Cloudonomics

Isnt the Cloud a Utility, and thus a Commodity?

The Pizza Problem

Great pizza!

My compliments to the electric utility.

2013 Telx. All Rights Reserved.

The Pizza Problem

Great online pizza ordering! My compliments to the cloud and network service providers.

Order Now

2013 Telx. All Rights Reserved.

Does IT Matter?

The Carr Argument

Ubiquity Commodity
Non-Strategic Minimize Spend

IT Spend and Profitability

8

Software & Internet

IT Spend as % of Revenues

7 6 5 4 3 2 1 0 5 7 9 11 13 15 17 19 21 23

Banking & Financial Education Healthcare Electronics Pro Services Telecom Insurance Transportation Mfg. Chem Retail / Wholesale Utilities Media & Entertainment Pharmaceuticals & Medical

Consumer Products Food Energy Mining & Construction

Operating Margin
Source: McKinsey, based on Gartner IT Metrics: IT Spending and Staffing Report, 2012; McKinsey Business Support Function (BSF) proprietary database

IT Spend and Profitability

8

Software & Internet

7

Relative Computerization

6 5 4 3 2 1 0 5 7 9 11 13 15 17 19 21

Banking & Financial Education Healthcare Electronics Pro Services Telecom Insurance Transportation Mfg. Chem Retail / Wholesale Utilities Media & Entertainment Pharmaceuticals & Medical

Consumer Products Food Energy Mining & Construction

23

Five Forces +
Source: McKinsey, based on Gartner IT Metrics: IT Spending and Staffing Report, 2012; McKinsey Business Support Function (BSF) proprietary database

10

Range of IT Value

Existential

Increasing Value

Strategic

Operational

Tactical
Source: Joe Weinman, Cloudonomics

Falsifying the Hypothesis

Source: Steve Jurvetson, Wikipedia and http://flickr.com/photos/jurvetson/157722937/

12

Explicit Core Search Share

EOY 2010 Google Sites 66.6%

EOY 2011 65.9%

EOY 2012 66.7%

Microsoft Sites
Yahoo! Sites Ask Network AOL

12.0%
16.0% 3.5% 1.9%

15.1%
14.5% 2.9% 1.6%

16.3%
12.2% 3.0% 1.8%

Source: comScore Explicit Core Search Share Reports, 2011-2013

Re-Imagination

Source: Mary Meeker/KPCB, Internet Trends, Stanford-Bases Kickoff, December 3, 2012

Creative Destruction
Destruction
Edited Print Encyclopedia Text Messaging

Creation
Crowdsourced Wikipedia Over the Top WhatsApp

Circuit-Switched Voice
Matchmakers Film Photography Books CDs Waving Down Cabs Cash Registers

Skype, Google Voice

Match.com Digital / Instagram eBooks iTunes / Pandora / Spotify Uber Square

Based in Part on Mary Meeker/KPCB, Internet Trends, D10 Conference, May 30, 2012

Decision Making

Data driven decision making correlates with higher productivity, market value, ROE, and asset utilization.

Source: Brynjolfsson, Hitt, and Kim, Strength in Numbers: How Does Data-Driven Decisionmaking Affect Firm Performance?, 2011

16

Firm Profitability

Fixed Effect Models*

Base Model IT Sales 1.912 Incl Sales, OPEX 1.837 0.054

OPEX
Advertising R&D R2 (overall) Observations Number of Firms 0.155 1.001 0.49

0.137
0.142 0.993 0.51 276 86

Source: Sunil Mithas, Ali Tafti, Indranil Bardhan, Jie Mein Goh, Information Technology and Firm Profitability, MIS Quarterly, March 2012. Used by Permission (*includes intercept, year, industry average Tobins q, Herfindahl index, and industry capital intensity)

17

How can the Cloud be Strategic?

The Experience Society

Societal Transformations Individual Transformations Experiences Services Products Commodities

Value

Complexity
Extending Pine & Gilmore, The Experience Economy

Value Disciplines

20

Value Disciplines

Operational Excellence

Product Leadership

Customer Intimacy

Accelerated Innovation
21

Source: Michael Treacy & Fred Wiersema, The Discipline of Market Leaders;Joe Weinman, Cloudonomics

Operational Excellence

Positions

Routes

Traffic

Source: Joe Weinman, Cloudonomics

22

Product Leadership

App Store Platform for Innovation Product-Service System

Source: Joe Weinman, Cloudonomics

23

Product Leadership

Emergency Assistance Real-Time Traffic Entertainment

24

Source: Joe Weinman, Cloudonomics

Product Leadership

Data Sharing Social Networking Goal Progress Community

25

Source: Joe Weinman, Cloudonomics

Customer Intimacy

Transactions

Recommendations

Frictionless Intimacy Business Analytics Promotions Customer Satisfaction

26

Source: Joe Weinman, Cloudonomics

Accelerated Innovation
Challenge

Solution Solutions

Source: Joe Weinman, Cloudonomics

27

The Networked Enterprise

When to Use the Cloud

Complementary Capabilities and Competencies

Consolidation
Cash Flow Capacity Continuity Checkpoints Chokepoints Context Celerity Customer Experience

Communications
Conversations, Connections, and Communities Congregations, Commons, and Collections Collaboration, Competition, and Crowdsourcing Commerce and Clearing Collaborative Consumption Coordination, Currency, Consistency, and Control

Cross-Device Access and Synchronization

Source: Joe Weinman, Cloudonomics

29

When Not to Use Cloud

Constant Demand

Custom
Classic Close Coupling Content Capture, Creation, and Consumption Cryptography Compression Caching

Covert
Continuity (of Local Services and Data)

Source: Joe Weinman, Cloudonomics

30

Product Life Cycle

Introduction Growth Maturity Decline

Revenue

Time
2013 Telx. All Rights Reserved.

Industry Rivals
Introduction Growth Maturity Decline

Industry Rivals

Time
2013 Telx. All Rights Reserved.

Industry Profit
Introduction Growth Maturity Decline

Industry Profit

Time
2013 Telx. All Rights Reserved.

Industry Profit
Introduction
Time Value Of Money

Growth
Higher Profit

Maturity

Decline

Industry Profit

Lower Marketing Costs Higher Market Share Lower Forecast Risk

Lower Development Expenses

Time
2013 Telx. All Rights Reserved.

Digital Business Intervals

Introduction Growth Maturity Decline

Time to Volume Time to Market

Infrastructure as a Service Platform as a Service Time
2013 Telx. All Rights Reserved.

Do Data Centers Use Too Much Energy?

Power, Pollution, and the Internet

Source: James Glanz, Power, Pollution, and the Internet, New York Times, September 22, 2012

Datacenter Electricity Use

Source: Jonathan Koomey, Growth in Data Center Electricity Use 2005 to 2010, August 1, 2011

Substitution

Energy Used

Source: Joe Weinman, Cloudonomics

39

Optimization

Energy Used

Source: Joe Weinman, Cloudonomics

40

Are Data Center Jobs Insignificant?

Jobs

Data Center Jobs

Digital Business Jobs

Indirect Jobs

Source: Joe Weinman, Why data centers have a big impact on the economy, GigaOM

42

What Is the Cloud?

NIST Definition of Cloud

Cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider nteraction.

On-demand self-service: A consumer can unilaterally provision computing capabilities, such as server time and network storage, as needed automatically without requiring human interaction with each services provider.

Broad network access: Capabilities are available over the network and accessed through standard mechanisms that promote use by heterogeneous thin or thick client platforms (e.g., mobile phones, laptops, and PDAs).

Resource pooling: The providers computing resources are pooled to serve multiple consumers using a multi-tenant model, with different physical and virtual resources dynamically assigned and reassigned according to consumer demand. There is a sense of location independence in that the customer generally has no control or knowledge over the exact location of the provided resources but may be able to specify location at a higher level of abstraction (e.g., country, state, or datacenter). Examples of resources include storage, processing, memory, network bandwidth, and virtual machines.

Rapid elasticity: Capabilities can be rapidly and elastically provisioned, in some cases automatically, to quickly scale out and rapidly released to quickly scale in. To the consumer, the capabilities available for provisioning often appear to be unlimited and can be purchased in any quantity at any time.

Measured Service: Cloud systems automatically control and optimize resource use by leveraging a metering capability at some level of abstraction appropriate to the type of service (e.g., storage, processing, bandwidth, and active user accounts). Resource usage can be monitored, controlled, and reported providing ransparency for both the provider and consumer of the utilized service.
Source: Peter Mell, Tim Grance, National Institute of Standards and Technology

Cloud = C.L.O.U.D.
Common
Penalty Cost Reduction to
1

for independent demands

Location-independent
Latency to
1

for n nodes, adjusted by

1cos(2) 1cos()

Online
Hub vs. P2P Benefit of Connections; Latency Penalty of Smart networks improvement on paths +
1 2 1 (+1) 1 4 3

45 128

Utility

Pure Cloud Better when > ; Hybrid Best with Breakeven at = /

on-Demand
Linear Benefit for = ; Exponential for = ;
Source: Joe Weinman, Cloudonomics
1

for Random Walk

45

Axiomatic Cloud Theory

A cloud is a structure: Space, a metric space = (,), Time, a measure space = (, , , <), A network = (, ), A set of states , each state comprising
A resource function : An allocation function : A location function : A pricing function : , a function space,

A transition mapping : , An initial state , = ( ) ; Satisfying five axioms: commonality, location independence, online (connectedness), utility (linear pricing), on demand
Source: Joe Weinman, Axiomatic Cloud Theory

46

Are Economies of Scale the Key to Cloud Benefits?

Cloud Cost Structures

Cost (/Price)

Consumer

Typical Well-Run Enterprise Enterprise

Mid-Size Cloud SP

Largest Cloud SP

Economies of Scale Diseconomies

Scale-Invariant Costs SG&A, Margin, Uncollectibles,

2013 Telx. All Rights Reserved.

48

The Math
1. U < 1 A * U * C * T <= P * U * C * T < P * 1 * C * T = P * C * T 2. U = 1 & A=P A * U * C * T = P * U * C * T = P * 1 * C * T = P * C * T 3. U = 1 & A < P A * U * C * T < P * U * C * T = P * 1 * C * T = P * C * T 4. 1 < U < ( P / A ) A * U * C * T < A * (P / A) * C * T = P * C * T 5. U > 1, & (Tp / T) < (1 / U) [(P - ) * T * C] + [ * Tp * C * U] < [(P - ) * T * C] + [ * T * C] 6. (TNZ / T) > (1 / U) & U > 1 * T * C < * TNZ * U * C 7. D uniform, U > 1 TotalCost = [F * T * C] + [ V * (V / P) * T * U * C] Min(TotalCost) when 0 = [T * C] * [0 1 + V * U / P] V / P = 1 / U

A = Average, P = Peak, C = Baseline Resource Cost, U = Utility Premium, T = Time, Tp = Peak Duration, TNZ = Non-Zero Baseline Duration, F = Fixed Capacity, V = Variable Capacity, D = Demand over Time

Source: Joe Weinman, Cloudonomics

49

The English
All other things being equal:

1. If cloud services cost less than enterprise IT, then

use them 2. If cloud services cost more than enterprise IT, then dont jump to conclusions, because if demand is spikier than the cloud is costly, a pure cloud solution will cost less than a dedicated one

3. If demand has any variation, a hybrid solution is optimal

Source: Joe Weinman, Cloudonomics

50

Hybrids

Users

Enterprise
Source: Joe Weinman, Cloudonomics

Cloud
51

Four Hybrids

Stateless Elasticity

BC/DR

Content/Service Delivery

Data Integration

Wont Only the Largest Providers Survive?

NBC Olympics: >10,000:1

2013 Alexa Internet. Used by Permission

54

Statistical Multiplexing

2013 Alexa Internet. Used by Permission

55

Statistical Multiplexing
1 0.8 1 0.8

0.6 0.4 0.2 0 1 0.8

0.6 0.4 0.2 0 1 0.8

25

0.6 0.4 0.2 0

100

0.6 0.4 0.2 0

Source: Joe Weinman, Cloudonomics

56

0.1 1

0.3

0.4

0.5

0.6

0.8

0.9

Relative Penalty Costs

0.2

0.7

Source: Joe Weinman, Cloudonomics

1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141 146 151 156 161 166 171 176 181 186 191 196 201

57

Which is Better, Consolidated Hyperscale or Dispersed?

Speed Revenue
Users really respond to speed Marissa Mayer Book and Merchandise: 100 ms = substantial revenue

Search: 10 30 results per page 400 ms 900 ms 20% drop in traffic 20% drop in revenue

James Hamilton, http://perspectives.mvdirona.com/2009/10/31/TheCostOfLatency.aspx

59

Application Evolution

Bandwidth Intensity

Latency Sensitivity
2013 Telx. All Rights Reserved. 60

Amdahls Law

Source: Joe Weinman, Cloudonomics

61

Cloud Response Time

Source: Joe Weinman, Cloudonomics

62

Cloud Response Time

n = QN 2P
0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

T=F+N+P
Single Entities

( )

Minimum Time for a given Quantity of resources

T = F + __ N + __ P n p
Multiplicity of Nodes and Processors

Priority of Dispersion

Source: Joe Weinman, Cloudonomics

63

How Important is On Demand

Linear Demand Growth

Demand D(t) = a+bt

Unserved Demand u

Resources R(t)

Provisioning Time tp

t1

Time
Source: Joe Weinman, Cloudonomics

t2
65

Exponential Demand

tp Demand D(t) = et

Unserved Demand
Resources R(t)=e(t-tp)

tp

t1

t2

Time
66

Source: Joe Weinman, Cloudonomics

Typical Demand Curve

Demand

Time
2013 Telx. All Rights Reserved.

Coarse Grained Billing

Capacity

Provisioning / Billing Increment

2013 Telx. All Rights Reserved.

Fine-Grained Billing

Capacity

Provisioning / Billing Increment

2013 Telx. All Rights Reserved.

Should Networks be Stupid or Smart?

Smart Networks
Software-Defined Networks / OpenFlow Network as a Service / Quantum Application-Aware Networking Programmability and Orchestration Internet Route Optimization Optically-Aware OpenFlow Extensions IRSCP, MPLS, Etc.
Visibility & Programmability

2013 Telx. All Rights Reserved.

71

Abstract Multi-Path Environment

Path 1 Path 2 Server 1 Server 2 Server 3 Path 3

Path 4

Server 4

Source: Joe Weinman, Better Together: Quantifying the Benefits of Smart Networks

72

Questions
Which approach yields the lowest response times?
Random Server-Based Load Balancing Joint Optimization

Are more options better? If so, how much better? What are the implications for distributed architectures and network intelligence?

2013 Telx. All Rights Reserved.

73

Server-Based vs. Network-Based

Source: Nick McKeown, How SDN Will Shape Networking, YouTube.com, Nikhil Handigol

74

Order Statistics Insights

Source: Joe Weinman, Better Together: Quantifying the Benefits of Smart Networks; Grateful Acknowledgement to Prof. H. N. Nagaraja, Ohio State University

75

Simulation Results

120 100 80 60 40 20 0 1 2 3 4 5 6 7 8 9 Joint Optimization Server Only Random

10

11

12

13

14

15

16

17

18

19

20

Source: Joe Weinman, Better Together: Quantifying the Benefits of Smart Networks, ComplexModels.com

Insights
More information is better than less More options are better than fewer Server-based load balancing trumps random

Joint optimization trumps server-based

Implementation requires network performance visibility and route control
Source: Joe Weinman, Better Together: Quantifying the Benefits of Smart Networks

77

Thank You
Questions?

2013 Telx. All Rights Reserved.

Welcome

2013 Telx. All Rights Reserved.