Sie sind auf Seite 1von 34

WORKING DRAFT

Last Modified 4/8/2009 9:13:54 PM Eastern Standard Time


Printed 1/28/2009 2:21:55 AM Central Standard Time

CONFIDENTIAL

Clearing the air on cloud


computing

Discussion Document
March 2009

CONFIDENTIAL AND PROPRIETARY


Any use of this material without specific permission of McKinsey & Company is strictly prohibited.
“ For I should not have rightly discovered things
celestial if I had not suspended the intellect and the
thought in a subtle form with its kindred air. But if,
being on the ground, I should never have
discovered them.

Aristophanes "The Clouds"

McKinsey & Company | 1


Context and objectives of this presentation

▪ Using “clouds” for computing tasks promises a revolution in IT similar to the birth
of the web and e-commerce
– Much lower cost
– Faster time to market
– Great opportunities for creating new sources of value

▪ While it has great potential, many of the claims being made about cloud computing have
lead some to the point of “irrational exuberance” and unrealistic expectations

▪ The purpose of this report is to focus the nascent cloud industry and its consumers on
setting realistic expectations by taking a “hype free” approach starting with the most
basic question of what a “cloud” actually is

▪ Our specific objectives include


– Proposing an industry-standard definition of cloud computing including what makes it unique
and exciting
– Identifying the types of customers that should be early adopters—those that can benefit from
existing and planned commercial offerings
– Understanding barriers that prevent large-scale adoption by corporations and government
entities that represent the bulk of IT spending

McKinsey & Company | 2


Key findings

“Cloud computing” is approaching the top of the Gartner Hype-cycle

Getting an industry “fix” on the definition of what a cloud


actually is would be a good first step and we propose one
for adoption here

Clouds already make sense for many small and medium-size


businesses, but technical, operational and financial hurdles
will need to be overcome before clouds will be used
extensively by large public and private enterprises

Rather than create unrealizable expectations for “internal clouds,” CIOs


should focus now on the immediate benefits of virtualizing server storage,
network operations, and other critical building blocks

Users, hardware vendors and service suppliers can take


specific steps to ensure the successful adoption of cloud
technology—and prevent it from getting stuck in the “trough of
disillusionment”

McKinsey & Company | 3


Cloud computing has tremendous promise

▪ Clouds can provide an


▪ Cloud computing can lower
IT barriers to innovation and
almost immediate
increase interoperability
access to hardware
between disjoint
resources
technologies. HP’s
▪ No upfront capital Creation of CloudPrint service is an
investments for users Faster time new value example
to market drivers

Lower upfront Easier to


IT costs scale-out
▪ Enterprises of all
▪ Pay-per-use billing sizes can seamlessly
dramatically reduces scale their services in
upfront costs correlation with client
demand

McKinsey & Company | 4


Cloud is quickly becoming top-of-mind in the IT world
2008 Gartner IT hype cycle

Since the publishing of


the 2008 cycle we have
seen a tremendous
The current U.S. economic
increase in coverage and woes will only drive more
talk about cloud enterprises to consider and
computing. adopt cloud offerings.
Spending on IT cloud
services will hit $42 billion
by 2012”

58 percent say cloud


computing will cause a
radical shift in IT and 47
percent say they're already
using it or actively
researching it”

SOURCE: Gartner; CIO magazine; IDC McKinsey & Company | 5


The tremendous buzz around clouds is being further fueled by intense
analyst and media speculation …

“noless influential than e-


business” (Gartner, 2008)

“ economic downturn, the appeal of


that cost advantage will be greatly
magnified" (IDC, 2008)
Not only is it faster and more flexible, it is
cheaper. […] the emergence of cloud models
radically alters the cost benefit
decision“ (FT Mar 6, 2009)
“Cloud computing achieves a quicker
return on investment “ “revolution, the biggest upheaval
(Lindsay Armstrong of salesforce.com, Dec 2008)
since the invention of the PC in the
1970s […] IT departments will have
little left to do once the bulk of
The economics are compelling, with business computing shifts […] into
business applications made three to five times the cloud” (Nicholas Carr, 2008)
cheaper and consumer applications
five to 10 times cheaper (Merrill Lynch,
May, 2008)

SOURCE: Press analysis McKinsey & Company | 6


… and seems to be following the familiar IT hype pattern (for CIOs at least)

Lack of interest in
Works well in theory The “Gold rush” technologies that work
Problem

Cloud computing has shown Initial excitement around a Cloud computing can divert
great promise for start-ups new technology grows rapidly IT departments’ attention
and pet projects for large into unrealistic expectations. from technologies that can
organizations. However, it is Early on, the main actually deliver sizeable
not ready to help with the big beneficiaries of emerging benefits; e.g., aggressive
challenges of big companies technologies are usually the virtualization
vendors, not the enterprise
users

HYPE

SOURCE: McKinsey & Company McKinsey & Company | 7


Key findings

“Cloud computing” is approaching the top of the Gartner Hype-cycle

Getting an industry “fix” on the definition of what a cloud


actually is would be a good first step and we propose one
for adoption here

Clouds already make sense for many small and medium-size


businesses, but technical, operational and financial hurdles
will need to be overcome before clouds will be used
extensively by large public and private enterprises

Rather than create unrealizable expectations for “internal clouds,” CIOs


should focus now on the immediate benefits of virtualizing server storage,
network operations, and other critical building blocks

Users, hardware vendors and service suppliers can take


specific steps to ensure the successful adoption of cloud
technology—and prevent it from getting stuck in the “trough of
disillusionment”

McKinsey & Company | 8


Why is it important to get a fix on clouds?

▪ Allows participants in cloud discussions to share a common


understanding of what is meant by clouds

▪ A rigorous definition allows CIOs to be analytical in their decision


making and to make more informed investment decisions

▪ Conversely, clarity also allows technology and service providers to


build meaningful product, marketing, and sales strategies that
translate into real value for their customers

▪ Allows an industry to begin moving forward with standards to


promote interoperability amongst cloud products

SOURCE: McKinsey & Company McKinsey & Company | 9


A recent study has found 22+ definitions of clouds

Many definitions
exist on cloud
computing

SOURCE: “A Break in the Clouds: Towards a Cloud Definition”, 2009; Vaquero, Rodero-Merino, Caceres, Linder McKinsey & Company | 10
Two notable attempts at a rigorous definition that captures key
characteristics

Does not distinguish


“Clouds are a large pool of easily usable and accessible virtualized cloud services from
resources (such as hardware, development platforms and/or services). clouds and does not
These resources can be dynamically reconfigured to adjust to a
variable load (scale), allowing also for an optimum resource utilization.
provide definitive
This pool of resources is typically exploited by a pay-per-use model in economic implications
which guarantees are offered by the Infrastructure Provider by means
of customized SLAs.” - Vaquero et al

“Cloud Computing refers to both the applications delivered as services Does not emphasize
over the Internet and the hardware and systems software in the abstraction of
datacenters that provide those services…The datacenter hardware and
software is what we will call a Cloud…Cloud computing has the
infrastructure explicitly
following characteristics

1. The illusion of infinite computing resources…


2. The elimination of an up-front commitment by Cloud users…
3. The ability to pay for use … as needed…” – UC Berkeley RAD Labs

SOURCE: “A Break in the Clouds: Towards a Cloud Definition”, 2009; Vaquero, Rodero-Merino, Caceres, Linder
“ Above the Clouds: A Berkeley View of Cloud Computing”, 2009; UC Berkeley Reliable Adaptive Distributed Systems Laboratory

McKinsey & Company | 11


How we have defined a cloud

Definition: Clouds are hardware-based services


offering compute, network and storage capacity where:

1 ▪ Hardware management is highly abstracted from the


buyer

2 ▪ Buyers incur infrastructure costs as variable OPEX

3 ▪ Infrastructure capacity is highly elastic (up or down)

SOURCE: McKinsey & Company McKinsey & Company | 12


Characteristics of clouds

Characteristic: Enterprises
incur no infrastructure capital
costs, just operational costs and
operational costs are incurred on
a pay-per-use basis, with no
contractual obligations

Definition: Clouds are hardware-based services


offering compute, network and storage capacity where:

1 ▪ Hardware management is highly abstracted from the


buyer

2 ▪ Buyers incur infrastructure costs as variable OPEX

3 ▪ Infrastructure capacity is highly elastic (up or down)

SOURCE: McKinsey & Company McKinsey & Company | 13


Characteristics of clouds

Characteristic: Enterprises
incur no infrastructure capital
costs, just operational costs and
operational costs are incurred on
a pay-per-use basis, with no
contractual obligations

Definition: Clouds are hardware-based services


offering compute, network and storage capacity where:

1 ▪ Hardware management is highly abstracted from the


buyer

2 ▪ Buyers incur infrastructure costs as variable OPEX

3 ▪ Infrastructure capacity is highly elastic (up or down)

Characteristic: Capacity can


be scaled up or down
dynamically, and immediately,
which differentiates from
traditional hosting service
providers

SOURCE: McKinsey & Company McKinsey & Company | 14


Characteristics of clouds

Characteristic: Enterprises
incur no infrastructure capital
costs, just operational costs and
operational costs are incurred on
a pay-per-use basis, with no
contractual obligations

Definition: Clouds are hardware-based services


offering compute, network and storage capacity where:

1 ▪ Hardware management is highly abstracted from the


buyer

2 ▪ Buyers incur infrastructure costs as variable OPEX

3 ▪ Infrastructure capacity is highly elastic (up or down)

Characteristic: Capacity can Characteristic: The


be scaled up or down underlying hardware
dynamically, and immediately, can be anywhere
which differentiates from geographically
traditional hosting service
providers

SOURCE: McKinsey & Company McKinsey & Company | 15


Characteristics of clouds
Characteristic: Architecture
specifics are abstracted. In
Characteristic: Enterprises addition, run in multi-tenancy
incur no infrastructure capital mode with multiple users
costs, just operational costs and accessing the infrastructure
operational costs are incurred on simultaneously
a pay-per-use basis, with no
contractual obligations

Definition: Clouds are hardware-based services


offering compute, network and storage capacity where:

1 ▪ Hardware management is highly abstracted from the


buyer

2 ▪ Buyers incur infrastructure costs as variable OPEX

3 ▪ Infrastructure capacity is highly elastic (up or down)

Characteristic: Capacity can Characteristic: The


be scaled up or down underlying hardware
dynamically, and immediately, can be anywhere
which differentiates from geographically
traditional hosting service
providers

SOURCE: McKinsey & Company McKinsey & Company | 16


Many cloud services are confused as clouds

1
Hardware management is
highly abstracted from the
buyer

Clouds Cloud Services


A true cloud has to A cloud service only
complies with two key
comply with all three
key requirements. It has
2 requirements: It is a
to abstract the service where the
Buyers incur infrastructure underlying infrastructure
underlying hardware
costs as variable OPEX is abstracted and can
from the buyer, be
elastic in scaling to scale elastically. It
demand and bill buyers could run on top of a
on a pay-per-use basis cloud, although it is not
required to (e.g., Saas)
3
Infrastructure capacity is
highly elastic (up or down)

SOURCE: McKinsey & Company McKinsey & Company | 17


Examples of clouds and cloud services

1
Hardware management is
highly abstracted from the
buyer

Cloud examples Cloud services


examples
2
Buyers incur infrastructure
costs as variable OPEX

3
Infrastructure capacity is
highly elastic (up or down)

SOURCE: McKinsey & Company McKinsey & Company | 18


Key findings

“Cloud computing” is approaching the top of the Gartner Hype-cycle

Getting an industry “fix” on the definition of what a cloud


actually is would be a good first step and we propose one
for adoption here

Clouds already make sense for many small and medium-size


businesses, but technical, operational and financial hurdles
will need to be overcome before clouds will be used
extensively by large public and private enterprises

Rather than create unrealizable expectations for “internal clouds,” CIOs


should focus now on the immediate benefits of virtualizing server storage,
network operations, and other critical building blocks

Users, hardware vendors and service suppliers can take


specific steps to ensure the successful adoption of cloud
technology—and prevent it from getting stuck in the “trough of
disillusionment”

McKinsey & Company | 19


Cloud offerings currently are most attractive for small and medium-sized
enterprises
Clouds are very cost effective for SMEs …
Total CPU/month cost

… and most customers of


Price for clouds are small businesses
public
cloud
consumers

Size
SME Large
enterprise

SOURCE: Amazon Web Site; McKinsey & Company McKinsey & Company | 20
There are significant hurdles to the adoption of cloud services
by large enterprises

Current cloud computing offerings are not cost-


1 Financial
effective compared to large enterprise data centers

Security and reliability concerns will have to be


2 Technical
mitigated and applications re-architected
Priority

Business perceptions of increased IT flexibility and


3 Operational
effectiveness will have to be properly managed

The IT supply and demand organizations will have to


4 Organizational
adapt to function in a cloud-centric world

McKinsey & Company | 21


Current cloud computing services are generally On-demand Windows

not cost effective for larger enterprises Pre-pay Linux

EC2 monthly CPU equivalent price options (virtual cores)


$

$180
▪ Most EC2 options are more
costly than TCO for a typical
data center
$140
▪ Enterprises could get lower
TCO through pre-pay
agreements—but only for
$100 Linux systems

$60
uneconomical The thin green line
TCA for typical data center1 around
economical $45/month for CPU equivalent
$20
Small Medium Large XL
EC2 Instance Size

1 Total Cost of Assets for “typical” data center: 10% utilization, $20M/MW for facility, $.1kW-hour, $14K/Server (2 CPU, 4 core)

SOURCE: Amazon Web Services; McKinsey & Company McKinsey & Company | 22
However, for smaller compute equivalents, today’s On-demand Windows

clouds may be attractive for some computing work Pre-pay Linux

EC2 monthly CPU equivalent price options (compute units2)


$ ▪ Not clear compute units are
the right measure of DC
$180 equivalency, as their
▪ Again, pre-pay
equivalent Linux years
is several
agreements
old show the
▪greatest
Again, promise, with some
pre-pay Linux
$140 EC2 options priced
agreements show the
significantly lower than
greatest promise, withTCO
some
forEC2
typical data priced
options center
$100 significantly lower than TCA
for typical data center

$60
uneconomical The thin green line
TCA for typical data center1 around
economical $45/month for CPU equivalent
$20

Small Medium Large XL


EC2 Instance Size
1 Total Cost of Assets for “typical” data center: 10% utilization, $20M/MW for facility, $.1kW-hour, $14K/Server (2 CPU, 4 core)
2 One EC2 Compute Unit provides the equivalent CPU capacity of a 1.0-1.2 GHz 2007 Opteron or 2007 Xeon processor

SOURCE: Amazon Web Services; McKinsey & Company McKinsey & Company | 23
The cost of cloud must come down significantly for DISGUISED CLIENT EXAMPLE

outsourcing a complete data center to make Labor costs


economic sense Non-labor costs

CPU/month TCO
Dollars

366

▪ Assumes migration of total


963 Windows and non-console Linux
capacity for entire data center
+144%
▪ Based on comparable hardware
configurations, current pricing for
cloud computing services is
150 significantly higher than the
CPU/month TCO achievable in
2701 data centers today.
107
▪ The key factor is that the majority
of servers that can be migrated
are Windows servers
432

CPU/month cost Cost of moving


of typical data to Amazon EC2
center2

1 Cost for comparable configuration – 75% of EC2 Large Standard Windows configuration
2 Typical CPU/month cost for 3GHz dual-core Xeon Windows-based servers
3 Estimated based on 10% labour savings from moving to a third-party cloud provider

SOURCE: Amazon Web Services; McKinsey & Company McKinsey & Company | 24
There is a 10-15% total infrastructure labor base DISGUISED CLIENT EXAMPLE

saving potential if moving whole data center to cloud Total starting FTEs
Total FTEs after migration

FTEs by role group

65
Application Support/Dev
63 ▪ Approx 10-15% of headcount can be
164 reduced by migrating capacity over to
Architecture 115 the cloud
45
Business Analysis
45 ▪ Most of this headcount comes from
52 facilities and touch labor roles
Client Services 52
131 ▪ Consequently, labor savings are
DB Admin 107
modest, but not insignificant…
257
Desk Top Support 257
20
Facilities 8 Total FTEs
87
Help Desk 87 1,704
673 1,448 -15%
IT Administrator
505
53
Management 53
40
Quality Assurance
40
1
Real Estate 1
116 Total starting FTEs Total FTEs after
Telecommunications 116 migration

SOURCE: Client Data; McKinsey & Company McKinsey & Company | 25


Many enterprise (necessarily or unnecessarily) set their SLAs uptimes at
99.99% or higher, which cloud providers have not yet been prepared to match

Amazon’s cloud outages receive a lot of exposure …

July 20, 2008 Failure due to stranded zombies,


lasts 5 hours
IT shops do not always deliver
Feb 15, 2008 Authentication overload leads to on their SLAs but their failures
two-hour service outage are less public and their
October 2007 Service failure lasts two days customers can’t switch easily
October 2006 Security breach where users could
see other users data

… and their current SLAs don’t match those of enterprises*

Amazon EC2 Amazon S3


99.95% 99.9% It is not clear that all
applications require such high
services levels

* SLAs expressed in Monthly Uptime Percentages

SOURCE: McKinsey & Company; Hyperbic Blog; UC Berkeley RADS Lab, CNet, eWeek; Amazon Web Site McKinsey & Company | 26
Key findings

“Cloud computing” is approaching the top of the Gartner Hype-cycle

Getting an industry “fix” on the definition of what a cloud


actually is would be a good first step and we propose one
for adoption here

Clouds already make sense for many small and medium-size


businesses, but technical, operational and financial hurdles
will need to be overcome before clouds will be used
extensively by large public and private enterprises

Rather than create unrealizable expectations for “internal clouds,” CIOs


should focus now on the immediate benefits of virtualizing server storage,
network operations, and other critical building blocks

Users, hardware vendors and service suppliers can take


specific steps to ensure the successful adoption of cloud
technology—and prevent it from getting stuck in the “trough of
disillusionment”

McKinsey & Company | 27


Large enterprises can achieve server utilization rates similar to those
cloud providers are achieving from their platforms …
Average server utilization rates

Best-in-class with
+280% 38
cloud (Google)

Best-in-Class with
+250% 35
aggressive virtualization

Generally achievable with


+150% 25 ▪ Most of the gains are achievable
aggressive virtualization
through standard virtualization
(~250% vs ~280%)
Quickly achievable +80% 18 ▪ However, even if cloud technolo-
gies were well implemented, most
companies would not achieve best-
Typical Utilization 10 in-class due to scale constraints
▪ Public clouds have a great deal
more flexibility to drive higher
utilization rates due to greater
variation in capacity requirements,
and the use of standard platforms,
which allows for better matching of
resources

SOURCE: McKinsey & Company;Google/ Barroso & Holzle McKinsey & Company | 28
… and, by adopting data center best practices, can ILLUSTRATIVE

drive down server TCO by more than 50% Key value drivers

Total monthly cost per CPU equivalent


$, including labor

Typical CPU
72.00
equivalent cost1

Drive utilization levels to 25% by leveraging virtualization,


21.50 1 Virtualization stacking, and consolidation

TCO Drive down demand by 20-30% by creating TCO transparency


2.00 2 transparency at the business level

Improve labor efficiency by over a third


19.00 3 Lean

By standardizing on platform, improve facility efficiency 20%


1.50 4 Service catalog and improve IT efficiency 10%, drive down IT capex 10%

Achievable CPU
28.00
Equiv. Cost
-61%

1 TCO for “typical” data center: 10% utilization, $20M/MW for facility, $14K/Server (2 CPU, 4 core), PUE of 2.0, $.1/kW-hour, facility utilization
of 65%, 30-40 Server/FTE ration, $100K/year/FTE, includes IT support and facility support labor

SOURCE: McKinsey & Company McKinsey & Company | 29


CIOs should be working the cloud technology stack bottom up; cloud
providers top down

Cloud provider technology stack Description

Billing ▪ Publicly-announced private


cloud computing environments
▪ Allows for dynamic provisioning of
Cloud Layer

Pricing are not by definition true


virtual machines clouds because enterprises
▪ Provides seamless failover capabilities can’t avoid capital
Disaster recovery
across geographic regions expenditures
Mobility ▪ Provides pricing and billing
functionality ▪ However, enterprises can
realize significant financial
Provisioning benefit by implementing
aggressive virtualization
▪ Contains hypervisor and virtual programs
Virtualization machines running in hypervisors
▪ When thinking about clouds,
layer ▪ Also contains OS instances CIOs should think from the
bottom of the stack up,
whereas providers of cloud
▪ Highly standardized IT hardware (e.g.,
services should take a top-
servers, network equipment, storage)
IT platform down approach
▪ OS that sits on top of hardware

Publicly-announced private
▪ The actual data center facilities clouds are essentially an
Facilities ▪ Highly efficient and standardized aggressive virtualization
across regions program on top of the
traditional enterprise IT stack

SOURCE: McKinsey & Company McKinsey & Company | 30


Key findings

“Cloud computing” is approaching the top of the Gartner Hype-cycle

Getting an industry “fix” on the definition of what a cloud


actually is would be a good first step and we propose one
for adoption here

Clouds already make sense for many small and medium-size


businesses, but technical, operational and financial hurdles
will need to be overcome before clouds will be used
extensively by large public and private enterprises

Rather than create unrealizable expectations for “internal clouds,” CIOs


should focus now on the immediate benefits of virtualizing server storage,
network operations, and other critical building blocks

Users, hardware vendors and service suppliers can take


specific steps to ensure the successful adoption of cloud
technology—and prevent it from getting stuck in the “trough of
disillusionment”

McKinsey & Company | 31


Avoiding the trough of disillusionment will require appropriate action
from all players in the cloud computing arena
▪ Develop an overall strategy for XaaS based on solid business cases not “cloud for the
Cloud users sake of cloud”
(e.g. CIOs, CTOs) ▪ Use modular design in all new and re-architected software to minimize costs when it
Stakeholder groups

comes time to migrate to cloud


▪ Set up Cloud CIO Council to advise industry

Cloud tools and ▪ Develop improved security standards to allay fears of client base
infrastructure ▪ Implement technologies that will allow for fine grain billing and management across a
cluster of compute devices
(e.g. HP, IBM)

Cloud providers ▪ In the near term, focus on small and medium sized businesses
(e.g. Amazon, ▪ Work on improving uptime rates into the 99.99% range
Azure) ▪ Continue to drive down prices through scale/innovation to increase potential market

The 2000 dot-com bubble provides an extreme example of the dangers of investing in hype…
5000
▪ Huge investments were made based largely on
4000 NASDAQ hype
Composite
3000 ~-80%
▪ Eventually, the inability to generate profits led to
2000
the collapse of most of the boom-time dot-coms

1000 ▪ From peak to trough, the NASDAQ-composite lost


~ 80% of it’s value
1994 1996 1998 2000 2002 2004 2006

SOURCE: McKinsey & Company McKinsey & Company | 32


Further information:

Will Forrest
william_forrest@mckinsey.com
+1 (312) 551-3975

ayewill #mckclouds

Media inquiries
Charlie Barthold
charles_barthold@mckinsey.com
+1 (203) 977-6915

McKinsey & Company | 33

Das könnte Ihnen auch gefallen