Cloud computing dummies phần 3
- 33 trang
- file .pdf
44 Part I: Introducing Cloud Computing
✓ You might decide to use Platform as a Service to limit the capital
expenses needed to develop a new application.
✓ Another starting point might be to add Software as a Service to analyze
what the market is saying about your products and any possible acquisi-
tion targets.
✓ Some organizations might have the need for a Business Process as a
Service (such as a supply chain service on demand) that could support
testing a new line of business.
Assessing Your Expense Structure
One of the most important tasks when preparing for the cloud: Assessing your
cost structure (for example, how much you’re spending on supporting existing
hardware, software, networking services). How can you determine the cost
savings if you don’t know what you’re spending today? Also take potential
future costs into account.
Things may get fuzzy. You may sometimes want to use business services
offered by cloud application vendors. You may want to build some internal
service oriented architecture-based services that can live inside a cloud envi-
ronment. In some situations, it may save money to move a service such as
email, software testing, or storage to a cloud, because the costs of perform-
ing the service internally are so much higher. In other situations, the costs
for implementing a key application in the cloud may be much more expensive
than running it internally.
Chapter 21 explains more about cloud economics.
Checking Up on Rules and Governances
We recommend assessing your current IT and business governance situation
as you develop your cloud strategy. In some cases, governance and compli-
ance prohibit certain types of information from leaving the organization’s
internal environment. How good is your internal security today? If you’re con-
sidering a cloud service provider, you need to be confident that the company
can support your security and governance needs with oversight and account-
ability. Examine the reports and documentation to support your oversight
requirements. Talk to the provider’s other customers to see how well it meets
its customers governance requirements.
Chapter 4: Developing Your Cloud Strategy 45
For example, you may want to leverage a third-party credit checking service
from the cloud. How well constructed is it? Does it conform to your com-
pany’s business rules?
Aside from security and privacy issues, you have a number of legal issues to
consider as well. For example, what happens to your application and data if
the cloud provider goes out of business? Who’s liable for lost information?
Does the provider guarantee uptime? What recourse do you have if the service
level agreement isn’t met?
Chapter 16 details governance issues.
Developing a Road Map
You must consider many things before developing a road map:
✓ The efficiency and effectiveness of your current data center
✓ Costs
✓ Risks
✓ Your organizational readiness
After you understand the issues and gaps, you can start designing your cloud
plan — the road map that outlines the following:
✓ What are the services that you need to support your business growth?
✓ How you will roll them out?
✓ When you will roll them out (or in, as it were)?
Don’t try to do everything at once with your cloud strategy. It probably makes
sense to roll out these services gradually so you can see the benefits and get
buy-in throughout your organization. Plus, starting cloud services step by step
can help you react quickly to business needs.
Even if you figure out all the technical requirements for leveraging the cloud as
part of your strategy, you still have to plan to communicate the action plan to
the business and the IT communities. Some people might consider the cloud a
threat because it will remove some tasks from the IT department. Business man-
agement will want to know that they have control over important business data.
For more details on your strategy action plan see Chapter 22.
46 Part I: Introducing Cloud Computing
You need to understand how your vendors track performance and security.
Don’t simply take their word for it and assume that everything is perfectly
fine. Even if the cloud vendor provides you with a slick dashboard, you should
have your own means of monitoring your content. You’re turning over some
key responsibility to a cloud provider, but the buck still stops with your orga-
nization. Plan carefully for controlling your assets in the cloud. Chapter 20
talks more about management from a cloud customer perspective.
Part II
Understanding the
Nature of the
Cloud
In this part . . .
W hat’s inside the cloud? In this part, we examine a
highly scaled computing environment. Because
that environment is front and center, we look at the tech-
nical foundation for this model, including workloads and
data services.
Chapter 5
Seeing the Advantages of the
Highly Scaled Data Center
In This Chapter
▶ Modeling a data center
▶ Location, location, location
▶ Powering things up
▶ Cooling things off
A s we discuss in Chapter 1, many company managers are demanding
that IT management transform their data centers into platforms that
can scale easily and effectively. Other managers are looking at the cloud plat-
form as a way to eliminate the high costs of running traditional data centers.
If you’re tasked with planning your cloud strategy, how do you do what’s
best for your organization? At first glance, it might seem obvious: Simply find
a cloud services provider, analyze how much it charges for the services you
need, and compare it to the costs of your own data center. It isn’t that simple.
✓ It’s unlikely that everything you do in your data center will be available
as a cloud service.
✓ Even if it is, it might not meet your specific needs.
Ultimately, cloud services are attractive because the cost is likely to be far
lower than providing the same service from your traditional data center, so
we think it will help if you understand why cloud data center costs are lower.
This economic factor applies to clouds whether they’re private or public.
50 Part II: Understanding the Nature of the Cloud
In fact, the cloud data center has two aspects:
✓ The costs of things that don’t depend directly on technology
✓ The costs of things that do
In this chapter, we take an in-depth look at the things that don’t depend
on technology and explain why the cloud data center has a significant cost
advantage.
Comparing Financial Damage:
Traditional versus Cloud
How much does a data center cost to run? It depends on these things:
✓ How big it is. How many virtual servers? Is the data center massive?
How much square footage; how many servers? Does it cost $5 million a
year to run?
✓ Where it is. How much does office space cost. What about cost of staff?
Is the data center close to inexpensive power sources?
✓ What it’s doing. Does the data center protect sensitive data? What is its
kind of business? What level of compliance must it adhere to?
Clearly, you have many ways to look at the situation.
Traditional data center
Although each data center is a little different, the average cost per year to
operate a large data center is usually between $10 million to $25 million.
Stranger than fiction
We didn’t make up the $10 million to $25 mil- images.businessweek.com/
lion number. In 2008, BusinessWeek Magazine ss/08/08/0804_cloudcomputing/1.
published an article called “Computing Heads htm). The magazine surveyed 11 different large
for the Clouds,” by Rachael King (http:// data centers throughout the United States.
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 51
Where’s the bulk of the money going? This might surprise you.
✓ 42 percent: Hardware, software, disaster recovery arrangements, unin-
terrupted power supplies, and networking. (Costs are spread over time,
amortized, because they are a combination of capital expenditures and
regular payments.)
✓ 58 percent: Heating, air conditioning, property and sales taxes, and
labor costs. (In fact, as much as 40 percent of annual costs are labor
alone.)
The reality of the traditional data center is further complicated because most
of the costs maintain existing (and sometimes aging) applications and infra-
structure. Some estimates show 80 percent of spending on maintenance.
Before you conclude that you need to throw out the data center and just move
to the cloud, know the nature of the applications and the workloads at the
core of data centers:
✓ Most data centers run a lot of different applications and have a wide
variety of workloads.
✓ Many of the most important applications running in data centers are
actually used by only a relatively few employees. For example, trans-
action management applications (which are critical to a company’s
relationship to customers and suppliers) might only be used by a few
employees.
✓ Some applications that run on older systems are taken off the market
(no longer sold) but are still necessary for business.
Because of the nature of these applications, it probably wouldn’t be cost
effective to move these environments to the cloud.
Cloud data center
In this case cloud data centers means data centers with 10,000 or more serv-
ers on site, all devoted to running very few applications that are built with
consistent infrastructure components (such as racks, hardware, OS, network-
ing, and so on).
What’s the key difference in the cost structure of a traditional data center and
a cloud data center? One of the most important factors is that cloud data cen-
ters aren’t remodeled traditional data centers.
52 Part II: Understanding the Nature of the Cloud
Cloud data centers are
✓ Constructed for a different purpose.
✓ Created at a different time than the traditional data center.
✓ Built to a different scale.
✓ Not constrained by the same limitations.
✓ Perform different workloads than traditional data centers.
Because of this design approach, the economics of a cloud data center are
significantly different.
To create a basis for analyzing this, we used figures on the costs of creating a
cloud data center described in a Microsoft paper titled “The Cost of a Cloud:
Research Problems in Data Center Networks” by Albert Greenberg, James
Hamilton, David A. Maltz, and Parveen Patel.
We took estimates for how much it cost to build a cloud data center and
looked at three cost factors:
✓ Labor costs were 6 percent of the total costs of operating the cloud data
center.
✓ Power distribution and cooling were 20 percent.
✓ Computing costs were 48 percent.
Of course, the cloud data center has some different costs than the traditional
data center (such as buying land and construction).
This explanation of costs is designed to give you an idea of where the differ-
ence between the traditional data center and the cloud data center are. The
upfront costs in constructing cloud data centers are actually spread across
hundreds of thousands of individual users. Therefore, after they’re con-
structed, these cloud data centers are well positioned to be profitable because
they support so many customers with a large number of servers executing a
single application.
Scaling the Cloud
From the provider’s point of view, the whole point of cloud computing is
to achieve economies of scale by managing a very large pool of computing
resources in a highly economic and efficient fashion.
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 53
A picture makes it a little clearer. Figure 5-1 shows a graph of the cost per
user of running just one software application using different kinds of com-
puter resources; this is charted against the number of users. We need to
emphasize that we’re talking about just one application — not even two or
three. In Figure 5-1, that one application runs in different computing environ-
ments, starting with inefficient dedicated servers all the way up to massively
scaled grids.
An important point to note is that the Y-axis of user populations is logarith-
mic. That means that the curve is much less steep than if we drew it on a pro-
portional scale of equal steps. If we drew it on a proportional scale, we’d need
miles of paper.
We deliberately didn’t put units on the X-axis. Instead, note the following:
✓ One end of the X-axis shows data center costs between $1–$50 per user
per annum. That reflects, for example, the prices that Google charges
for Google Apps or even the cost of providing free email (from Google,
Microsoft, or Yahoo, which is paid for by ads). The cost per user is
extremely low.
✓ The other end of the X-axis shows data center costs between $1,000–
$5,000 per user per annum. That might be the cost of, for example, pro-
viding a print server that’s almost always idle.
User
Population Scaling Out
Cloud
1,000,000,000 Computing
100,000,000 Massively
Scaled Grid
10,000,000
1,000,000 Large Grids
100,000 Grids
10,000 Mainframe
1,000 Large Unix Clusters
Efficient Servers
100
Figure 5-1: Virtual
10 Mixed Machines Inefficient
Cloud Workloads Servers
computing 1
economies
of scale.
$1-$50 p.a. Costs Per User $1000-$5000 p.a.
54 Part II: Understanding the Nature of the Cloud
Basically, on the left in Figure 5-1 you have very efficient use of computer
resources and, on the right, very inefficient use of resources.
Points on the line indicate the kind of computing resources that serve spe-
cific group sizes:
✓ Inefficient servers: This is a 1:1 user-to-server ratio (or close to 1). The
cost of managing a single server in a data center will be thousands of
dollars per year and this is as expensive as computing ever gets per
user.
✓ Virtual machines: Applications and user numbers that can’t use a whole
server get virtualized (split among several virtual servers). This is effi-
cient (making better use of underused servers), but also inefficient (vir-
tualization requires significant overhead, as does running the multiple
guest operating systems).
✓ Efficient servers (and small clusters): User populations from the hun-
dreds to 1,000 can be served reasonably efficiently with a single or multi-
ple servers if there’s only one application being run on a server; servers
can be highly efficient, yielding a relatively low cost per user.
✓ Mainframe and large Unix clusters: They’re shown separately on the
grid only for the sake of space. Both can handle very large database
applications from thousands to tens of thousands of users.
✓ Grids: From the hundreds of thousands to a million users, you’re in the
area where Software as a Service (SaaS) vendors such as Salesforce.com
operate. Business applications offered by SaaS vendors present a thorny
scaling problem because it’s a transactional database application. The main
Salesforce.com CRM application runs on a grid of about 1,000 computers.
✓ Large grids: Concurrent users above one million. Still a very heavy
workload and only possible via a scale-out (which lets a single workload
expand by using more of the identical inexpensive resources) approach
with a grid. Twitter and Linked-In are examples.
✓ Massively scaled grid: This is for user populations in the tens of millions.
Example: Each query on Google search is resolved by a purpose-built grid
of up to 1,000 servers; Google routes queries to many such grids. Yahoo
also has a massively scaled-out email system. It caters to more than 260
million users, of which tens of millions must be active at a time.
The dotted box in Figure 5-1 indicates the traditional domain and kinds of
resources of corporate computing. The same servers used in corporate envi-
ronments could be used just as easily in scaled-out arrangements, where
workloads aren’t at all mixed. The reduction in per-user costs doesn’t, at
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 55
the moment, come from using different computer equipment or different
operating systems: It comes from running a small number (or even just one)
workload and scaling it up as much as possible. That’s how cloud computing
reduces costs dramatically.
No corporation that runs a mixed workload is ever going to achieve cloud
computing’s economies of scale.
But how do massively scaled data centers manage to get their per-user costs
so very low? This becomes clear when you read about each area of data
costs in Chapter 21.
Comparing Traditional and
Cloud Data Center Costs
Before reading how to reduce data center costs, reread the traditional IT
costs statistics:
✓ Portion of IT budget used to maintain and run existing systems:
70–80 percent
✓ Portion of IT budget used to build and implement new capabilities:
20–30 percent
Compare traditional and cloud data centers in Table 5-1.
Table 5-1 A Comparison of Corporate and Cloud Data Centers
Traditional Corporate Data Center Cloud Data Center
Thousands of different Few applications (maybe even just one)
applications
Mixed hardware environment Homogeneous hardware environment
Multiple management tools Standardized management tools
Frequent application patching and Minimal application patching and updating
updating
Complex workloads Simple workloads
Multiple software architectures Single standard software architecture
56 Part II: Understanding the Nature of the Cloud
Looking at the table, it becomes clear that the cloud data center is much sim-
pler to organize and operate and, because it is simple, it scales well. In other
words, the larger you make it, the lower the costs per user are. In the next
section, we examine some of these costs and see where the efficiencies arise.
Examining labor costs and productivity
Labor costs depend on several things:
✓ Technology managing the data center: Even improving that technology
in a traditional corporate setting may reduce the cost of labor only a
small amount.
✓ In what environment someone works: The labor cost per person is
likely to be equivalent regardless of the data center type; the skills
requirement is the same. But that person’s productivity varies depend-
ing on the environment. Operating the scaled cloud data center is much
simpler.
The impact of this set of differences on labor costs is dramatic. Corporate
data centers usually have a ratio of operational staff to severs of around 1
person to 65 servers. In cloud data centers, that ratio is more like 1 person
to 850 servers, and we’ve even come across better ratios than that. This is a
10-to-1 improvement in the productivity of labor (or possibly more — maybe
going as high as 20 to 1).
Wondering where you are
The traditional setup’s 58 percent costs depend a lot on location:
✓ Electricity fees
✓ Local taxes
✓ Labor costs
Compare a data center in North Carolina with one in New York (keeping in
mind no two data centers have the same software workloads). Better to con-
sider technology costs separately and see where economies arise, which we
do in the following sections.
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 57
Electric power
Computers have been using more electricity in recent years and, at 7 percent
of corporate data centers’ costs (including heating and cooling), the cost is
significant. Cloud data centers use even more: Electricity costs hover around
12 percent.
Cloud data centers can do the following:
✓ Put the data center where the cheap power is. Electricity fluctuates in
price from year to year and costs are difficult to control.
✓ Negotiate a discounted power contract with its power company.
Cloud data centers, by their level of usage, fall into the least expensive
category.
If a cloud data center is contemplating building a data center,it can
negotiate a long-term deal for an even deeper discount than industrial
usage gives them. Put the data center very close to the power plant and
bargain for a lower cost supply based on these points:
• Distance from the power station (because less electrical power is
lost in transit).
• Minimal power interruption from electrical storms (if you have a
private circuit direct to the power station).
Outsourcing
Because power is so critical to the cloud data ✓ Oil prices change, which can cause cost
center, organizations have to consider the fluctuation.
availability and cost of energy sources as they
✓ Liquified natural gas (LNG) suffers from the
would any primary data center resource.
same changing fuel prices as oil.
Electricity sources include the following:
✓ Coal is more stable in price, but not green.
✓ Hydroelectric is generally expensive when
✓ Nuclear is inexpensive to run but expensive
it has to travel far to customers, but other-
to build and gain approval
wise it’s usually cheap and can be the ideal
source of power for a data center.
58 Part II: Understanding the Nature of the Cloud
Other location costs
Other location related costs when building a new data center include the
following:
✓ Land costs: The days of siting data centers in skyscrapers in Manhattan
are over. Better to use cheap land with low property taxes. There are
exceptions, of course. For example, in algorithmic financial trading,
latency lost due to networking (communications) distance directly
impacts revenue.
✓ Building costs: A designed-entirely-as-a-data-center building is a must.
• Heat management is the overriding priority, so building out almost
certainly makes more sense than building up. Cool geographical
areas may make more sense than hotter ones.
• Safety is another important consideration. Data centers need to be
electrically safe, secure, and fireproof.
✓ Staff: Although staff costs are very low for the cloud data center, as a
percentage of the whole, location in areas (or even countries) where
staff costs are low can further reduce staff costs.
✓ Investment incentives and taxation: Many areas of the world, including
states in the United States, welcome inward investment and help finance
it with very generous tax exemptions and cash incentives. Take advan-
tage of these opportunities when you find them.
In the next chapter, we examine technology costs, which also favor the cloud
data center in many ways. The simple fact is that data centers as they exist
now, in the enterprise, are a cottage industry that’s going to change in the
coming years by the mass-production efficiencies of cloud data centers.
Chapter 6
Exploring the Technical
Foundation for Scaling
Computer Systems
In This Chapter
▶ Comparing traditional data centers to clouds
▶ Achieving economies of scale
▶ Saving money via the bottom line
I n Chapter 5, we contrast the non-technology operational costs of the tra-
ditional data center with those of the cloud data center (electricity, cool-
ing, space, and so on). In this chapter, we contrast technology costs between
the traditional data center and the cloud data center.
We divided into four areas the places where IT spends money:
✓ Hardware, including servers, storage, and so on
✓ A power supply for those systems and how to keep them from overheating
✓ Networking and communications equipment so the systems can
interoperate
✓ Electricity to support the overall data center
Some elements are more expensive than others. In Chapter 5, we look at two
reports that detail the costs of running traditional and cloud data centers.
Using this same set of numbers, we calculated the costs of the areas. The
results are quite interesting. The greatest expense in the traditional data
center is server and storage hardware, which accounts for 36 percent of the
amortized costs. The second biggest expense? Power distribution and cool-
ing. Amortized over a year, power and cooling are 20 percent of the total
60 Part II: Understanding the Nature of the Cloud
expenses. Both networking and electricity each add 12 percent to the total
expense number per year. Add hardware and its supporting power and cool-
ing, and you have 56 percent of the technology related costs.
We discuss electricity costs in the previous chapter, but only from the per-
spective of arranging for an inexpensive supply. In this chapter, we take on
the issue of using that electricity efficiently.
Server-ing Up Some Hardware
Although we’d like to tell you that costs are static, clearly they aren’t. Costs
for your data center hardware will vary dramatically depending on the type of
workloads you support.
Data storage is an excellent example of this variation. If a data center is feed-
ing video to the Internet from a vast video library (like YouTube does) the
storage requirements are huge. However, storing short text messages (as
Twitter does) doesn’t require a lot of space. Indeed, Twitter doesn’t even
store its billions of messages indefinitely. The YouTube library, on the other
hand, just keeps on growing.
Tradition! versus clouds
What does this mean when you look at the differences in the costs of hard-
ware between the traditional data center and the cloud data center? Look at
a snapshot of each:
✓ Tradition: In a traditional data center, IT management has a structured
process for purchasing hardware. Each year they talk to business units,
determine what new applications to add or expand, and work with ven-
dors on the procurement process. In addition, most IT organizations
refresh their hardware on a regular basis to make sure that things run
smoothly and old systems are retired before they cause problems.
✓ Cloud: When a business is creating a cloud data center (either a private
one inside the firewall or a service provider) the process of procuring
systems is very different. Because the cloud supports very different
workloads, IT management doesn’t buy traditional hardware. Rather, IT
management might go directly to an engineering company that designs
the system boards and networking switches for them, and then take the
contract to a manufacturer to have them build the precise hardware
that they want.
Chapter 6: Exploring the Technical Foundation for Scaling Computer Systems 61
The bottom line is that the cloud data center is well suited to buying precisely
what you need in a very economical manner. In contrast, the traditional data
center doesn’t have the same economies of scale.
We aren’t being critical of the server products that are built and delivered by
big computer manufacturers. Such engineering is difficult to criticize in its nat-
ural context. All such servers, whether mainframes or cheap commodity
server boards, are designed for general circumstances of typical customers.
It’s just really unlikely that the requirements of a cloud center are anywhere
close to typical.
Considering cloud hardware
When your company is establishing a cloud data center, think about the
hardware elements in a different way. The following sections summarize
considerations.
Cooling
Cloud data centers have the luxury of being able to engineer the way systems
(boards, chips, and more) are cooled. When systems are cooled via air condi-
tioning, they require tremendous amounts of power. However, purpose-built
cloud data centers can be engineered to be cooled by water, for example
(which is 3,000 times more efficient than air in cooling equipment).
CPU, memory, and local disk
Traditional data tends to be filled with a lot of surplus equipment (either
to support unanticipated workloads or because an application or process
wasn’t engineered to be efficient). Surplus memory, CPUs, and disks take up
valuable space and, of course, they need to be cooled. The cloud data center
typically supports self-service provisioning of resources so capacity is added
only when you need it.
Data storage and networking
Data storage and networking need to be managed collectively if they’re going
to be efficient. This problem has complicated the way the traditional data
centers have been managed, and has forced organizations to buy a lot of
additional hardware and software. The cloud data center can be engineered
to overcome this problem. The cloud knows where its data needs to be
because it is so efficient in the way it manages workloads. The cloud actually
is engineered to manage data efficiently.
62 Part II: Understanding the Nature of the Cloud
Redundancy
Data centers must always move data around the network for backup and
disaster recovery. Traditional data centers support so many different work-
loads that many approaches to backup and recovery have to be taken. This
makes backing up and recovering data complicated and expensive. The
cloud, in contrast, is designed to handle data workloads consistently. For
example, in a cloud data center you can establish a global policy about how
and when backups will be handled. This can be then handled in an automated
manner, reducing the cost of handling backup and recovery.
Software embedded within the data center
We talk a lot about software in the context of applications, but a considerable
amount of software is linked at a systems level. This type of system level soft-
ware is a big cost in the traditional data center simply because there are so
many more workloads with so many operating systems and related software
elements.
As you know, cloud data centers have fewer elements because they have
simpler workloads. There are some differences in how software costs are
managed depending on the type of cloud model. Cloud providers understand
these costs well and design their offerings to maximize revenue. It will help
you understand pricing by understanding the cost factors for each of the
models.
The following gives you a sense of the difference between IaaS, PaaS, and
SaaS when it comes to embedded software costs:
✓ An Infrastructure as a Service (IaaS) operation is likely to have higher
software costs because although it provides only an environment for
running applications, it has to build that environment according to
equivalent environments in corporate data centers. Therefore, the IaaS
vendor has to spend a lot of resources on management and security
software in addition to the operating systems. See Chapter 10 for more
about IaaS.
✓ With a Platform as a Service (PaaS) operation, the provider delivers a
full software stack. To reduce cost, the PaaS vendor is likely to provide
a software stack consisting of proprietary components. The licens-
ing costs may be lower for IaaS than the PaaS environment because
the operator is likely to force the use of specific software products.
However, the PaaS vendor must maintain and support the software stack
it provides. See Chapter 11 for more about PaaS.
Chapter 6: Exploring the Technical Foundation for Scaling Computer Systems 63
✓ With Software as a Service (SaaS), the SaaS vendor provides a propri-
etary application as its value to customers. While the vendor invests in
this software, it typically relies on partners to support many of the other
functions. These vendors also take advantage of open-source compo-
nents. See Chapter 12 for more about SaaS.
Open-source dynamic
The cloud is an economic and business model as much as a technology model.
It isn’t surprising, then, that open-source software is an important element
for almost all cloud providers. Some of it is very high quality and nearly all of
it can be used for no license fee, as long as you obey the restrictions of the
associated license.
Open-source software has already become a business factor in the Internet
service provider (ISP) business, with most ISPs providing an easily installed,
highly functional software stack for building Web sites. Many cloud providers
take open-source software as a foundation and customize it to optimize sup-
port for their workloads.
The other software area that impacts costs is the way operating systems
are handled in the data center. Under traditional operation, an OS has many
background processes running. All such processes have a function and quite
a few of them run by default, whether you need them or not. Some of them
are keeping logs, some are handling messages from the network, some fire off
scheduled jobs, some handle printing, some provide directory services, and
so on. They all sit there happily chewing up CPU cycles. None of them should
be there unless they have a specific role to play.
In a traditional environment, no one would think of deleting useful background
processes, but nothing superfluous should run in an environment that prizes
efficient resource usage. Not only that, but if you’re running a cloud data
center, you may be interested in rewriting some of these tasks because you
need them to run slightly differently. That’s why open source plays a large
role in cloud operations.
Economies of Scale
We spend a lot of time in this chapter saying why the economics of the cloud
are so different than that of the traditional data center. Of course, not every
workload is right for the cloud.
✓ You might decide to use Platform as a Service to limit the capital
expenses needed to develop a new application.
✓ Another starting point might be to add Software as a Service to analyze
what the market is saying about your products and any possible acquisi-
tion targets.
✓ Some organizations might have the need for a Business Process as a
Service (such as a supply chain service on demand) that could support
testing a new line of business.
Assessing Your Expense Structure
One of the most important tasks when preparing for the cloud: Assessing your
cost structure (for example, how much you’re spending on supporting existing
hardware, software, networking services). How can you determine the cost
savings if you don’t know what you’re spending today? Also take potential
future costs into account.
Things may get fuzzy. You may sometimes want to use business services
offered by cloud application vendors. You may want to build some internal
service oriented architecture-based services that can live inside a cloud envi-
ronment. In some situations, it may save money to move a service such as
email, software testing, or storage to a cloud, because the costs of perform-
ing the service internally are so much higher. In other situations, the costs
for implementing a key application in the cloud may be much more expensive
than running it internally.
Chapter 21 explains more about cloud economics.
Checking Up on Rules and Governances
We recommend assessing your current IT and business governance situation
as you develop your cloud strategy. In some cases, governance and compli-
ance prohibit certain types of information from leaving the organization’s
internal environment. How good is your internal security today? If you’re con-
sidering a cloud service provider, you need to be confident that the company
can support your security and governance needs with oversight and account-
ability. Examine the reports and documentation to support your oversight
requirements. Talk to the provider’s other customers to see how well it meets
its customers governance requirements.
Chapter 4: Developing Your Cloud Strategy 45
For example, you may want to leverage a third-party credit checking service
from the cloud. How well constructed is it? Does it conform to your com-
pany’s business rules?
Aside from security and privacy issues, you have a number of legal issues to
consider as well. For example, what happens to your application and data if
the cloud provider goes out of business? Who’s liable for lost information?
Does the provider guarantee uptime? What recourse do you have if the service
level agreement isn’t met?
Chapter 16 details governance issues.
Developing a Road Map
You must consider many things before developing a road map:
✓ The efficiency and effectiveness of your current data center
✓ Costs
✓ Risks
✓ Your organizational readiness
After you understand the issues and gaps, you can start designing your cloud
plan — the road map that outlines the following:
✓ What are the services that you need to support your business growth?
✓ How you will roll them out?
✓ When you will roll them out (or in, as it were)?
Don’t try to do everything at once with your cloud strategy. It probably makes
sense to roll out these services gradually so you can see the benefits and get
buy-in throughout your organization. Plus, starting cloud services step by step
can help you react quickly to business needs.
Even if you figure out all the technical requirements for leveraging the cloud as
part of your strategy, you still have to plan to communicate the action plan to
the business and the IT communities. Some people might consider the cloud a
threat because it will remove some tasks from the IT department. Business man-
agement will want to know that they have control over important business data.
For more details on your strategy action plan see Chapter 22.
46 Part I: Introducing Cloud Computing
You need to understand how your vendors track performance and security.
Don’t simply take their word for it and assume that everything is perfectly
fine. Even if the cloud vendor provides you with a slick dashboard, you should
have your own means of monitoring your content. You’re turning over some
key responsibility to a cloud provider, but the buck still stops with your orga-
nization. Plan carefully for controlling your assets in the cloud. Chapter 20
talks more about management from a cloud customer perspective.
Part II
Understanding the
Nature of the
Cloud
In this part . . .
W hat’s inside the cloud? In this part, we examine a
highly scaled computing environment. Because
that environment is front and center, we look at the tech-
nical foundation for this model, including workloads and
data services.
Chapter 5
Seeing the Advantages of the
Highly Scaled Data Center
In This Chapter
▶ Modeling a data center
▶ Location, location, location
▶ Powering things up
▶ Cooling things off
A s we discuss in Chapter 1, many company managers are demanding
that IT management transform their data centers into platforms that
can scale easily and effectively. Other managers are looking at the cloud plat-
form as a way to eliminate the high costs of running traditional data centers.
If you’re tasked with planning your cloud strategy, how do you do what’s
best for your organization? At first glance, it might seem obvious: Simply find
a cloud services provider, analyze how much it charges for the services you
need, and compare it to the costs of your own data center. It isn’t that simple.
✓ It’s unlikely that everything you do in your data center will be available
as a cloud service.
✓ Even if it is, it might not meet your specific needs.
Ultimately, cloud services are attractive because the cost is likely to be far
lower than providing the same service from your traditional data center, so
we think it will help if you understand why cloud data center costs are lower.
This economic factor applies to clouds whether they’re private or public.
50 Part II: Understanding the Nature of the Cloud
In fact, the cloud data center has two aspects:
✓ The costs of things that don’t depend directly on technology
✓ The costs of things that do
In this chapter, we take an in-depth look at the things that don’t depend
on technology and explain why the cloud data center has a significant cost
advantage.
Comparing Financial Damage:
Traditional versus Cloud
How much does a data center cost to run? It depends on these things:
✓ How big it is. How many virtual servers? Is the data center massive?
How much square footage; how many servers? Does it cost $5 million a
year to run?
✓ Where it is. How much does office space cost. What about cost of staff?
Is the data center close to inexpensive power sources?
✓ What it’s doing. Does the data center protect sensitive data? What is its
kind of business? What level of compliance must it adhere to?
Clearly, you have many ways to look at the situation.
Traditional data center
Although each data center is a little different, the average cost per year to
operate a large data center is usually between $10 million to $25 million.
Stranger than fiction
We didn’t make up the $10 million to $25 mil- images.businessweek.com/
lion number. In 2008, BusinessWeek Magazine ss/08/08/0804_cloudcomputing/1.
published an article called “Computing Heads htm). The magazine surveyed 11 different large
for the Clouds,” by Rachael King (http:// data centers throughout the United States.
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 51
Where’s the bulk of the money going? This might surprise you.
✓ 42 percent: Hardware, software, disaster recovery arrangements, unin-
terrupted power supplies, and networking. (Costs are spread over time,
amortized, because they are a combination of capital expenditures and
regular payments.)
✓ 58 percent: Heating, air conditioning, property and sales taxes, and
labor costs. (In fact, as much as 40 percent of annual costs are labor
alone.)
The reality of the traditional data center is further complicated because most
of the costs maintain existing (and sometimes aging) applications and infra-
structure. Some estimates show 80 percent of spending on maintenance.
Before you conclude that you need to throw out the data center and just move
to the cloud, know the nature of the applications and the workloads at the
core of data centers:
✓ Most data centers run a lot of different applications and have a wide
variety of workloads.
✓ Many of the most important applications running in data centers are
actually used by only a relatively few employees. For example, trans-
action management applications (which are critical to a company’s
relationship to customers and suppliers) might only be used by a few
employees.
✓ Some applications that run on older systems are taken off the market
(no longer sold) but are still necessary for business.
Because of the nature of these applications, it probably wouldn’t be cost
effective to move these environments to the cloud.
Cloud data center
In this case cloud data centers means data centers with 10,000 or more serv-
ers on site, all devoted to running very few applications that are built with
consistent infrastructure components (such as racks, hardware, OS, network-
ing, and so on).
What’s the key difference in the cost structure of a traditional data center and
a cloud data center? One of the most important factors is that cloud data cen-
ters aren’t remodeled traditional data centers.
52 Part II: Understanding the Nature of the Cloud
Cloud data centers are
✓ Constructed for a different purpose.
✓ Created at a different time than the traditional data center.
✓ Built to a different scale.
✓ Not constrained by the same limitations.
✓ Perform different workloads than traditional data centers.
Because of this design approach, the economics of a cloud data center are
significantly different.
To create a basis for analyzing this, we used figures on the costs of creating a
cloud data center described in a Microsoft paper titled “The Cost of a Cloud:
Research Problems in Data Center Networks” by Albert Greenberg, James
Hamilton, David A. Maltz, and Parveen Patel.
We took estimates for how much it cost to build a cloud data center and
looked at three cost factors:
✓ Labor costs were 6 percent of the total costs of operating the cloud data
center.
✓ Power distribution and cooling were 20 percent.
✓ Computing costs were 48 percent.
Of course, the cloud data center has some different costs than the traditional
data center (such as buying land and construction).
This explanation of costs is designed to give you an idea of where the differ-
ence between the traditional data center and the cloud data center are. The
upfront costs in constructing cloud data centers are actually spread across
hundreds of thousands of individual users. Therefore, after they’re con-
structed, these cloud data centers are well positioned to be profitable because
they support so many customers with a large number of servers executing a
single application.
Scaling the Cloud
From the provider’s point of view, the whole point of cloud computing is
to achieve economies of scale by managing a very large pool of computing
resources in a highly economic and efficient fashion.
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 53
A picture makes it a little clearer. Figure 5-1 shows a graph of the cost per
user of running just one software application using different kinds of com-
puter resources; this is charted against the number of users. We need to
emphasize that we’re talking about just one application — not even two or
three. In Figure 5-1, that one application runs in different computing environ-
ments, starting with inefficient dedicated servers all the way up to massively
scaled grids.
An important point to note is that the Y-axis of user populations is logarith-
mic. That means that the curve is much less steep than if we drew it on a pro-
portional scale of equal steps. If we drew it on a proportional scale, we’d need
miles of paper.
We deliberately didn’t put units on the X-axis. Instead, note the following:
✓ One end of the X-axis shows data center costs between $1–$50 per user
per annum. That reflects, for example, the prices that Google charges
for Google Apps or even the cost of providing free email (from Google,
Microsoft, or Yahoo, which is paid for by ads). The cost per user is
extremely low.
✓ The other end of the X-axis shows data center costs between $1,000–
$5,000 per user per annum. That might be the cost of, for example, pro-
viding a print server that’s almost always idle.
User
Population Scaling Out
Cloud
1,000,000,000 Computing
100,000,000 Massively
Scaled Grid
10,000,000
1,000,000 Large Grids
100,000 Grids
10,000 Mainframe
1,000 Large Unix Clusters
Efficient Servers
100
Figure 5-1: Virtual
10 Mixed Machines Inefficient
Cloud Workloads Servers
computing 1
economies
of scale.
$1-$50 p.a. Costs Per User $1000-$5000 p.a.
54 Part II: Understanding the Nature of the Cloud
Basically, on the left in Figure 5-1 you have very efficient use of computer
resources and, on the right, very inefficient use of resources.
Points on the line indicate the kind of computing resources that serve spe-
cific group sizes:
✓ Inefficient servers: This is a 1:1 user-to-server ratio (or close to 1). The
cost of managing a single server in a data center will be thousands of
dollars per year and this is as expensive as computing ever gets per
user.
✓ Virtual machines: Applications and user numbers that can’t use a whole
server get virtualized (split among several virtual servers). This is effi-
cient (making better use of underused servers), but also inefficient (vir-
tualization requires significant overhead, as does running the multiple
guest operating systems).
✓ Efficient servers (and small clusters): User populations from the hun-
dreds to 1,000 can be served reasonably efficiently with a single or multi-
ple servers if there’s only one application being run on a server; servers
can be highly efficient, yielding a relatively low cost per user.
✓ Mainframe and large Unix clusters: They’re shown separately on the
grid only for the sake of space. Both can handle very large database
applications from thousands to tens of thousands of users.
✓ Grids: From the hundreds of thousands to a million users, you’re in the
area where Software as a Service (SaaS) vendors such as Salesforce.com
operate. Business applications offered by SaaS vendors present a thorny
scaling problem because it’s a transactional database application. The main
Salesforce.com CRM application runs on a grid of about 1,000 computers.
✓ Large grids: Concurrent users above one million. Still a very heavy
workload and only possible via a scale-out (which lets a single workload
expand by using more of the identical inexpensive resources) approach
with a grid. Twitter and Linked-In are examples.
✓ Massively scaled grid: This is for user populations in the tens of millions.
Example: Each query on Google search is resolved by a purpose-built grid
of up to 1,000 servers; Google routes queries to many such grids. Yahoo
also has a massively scaled-out email system. It caters to more than 260
million users, of which tens of millions must be active at a time.
The dotted box in Figure 5-1 indicates the traditional domain and kinds of
resources of corporate computing. The same servers used in corporate envi-
ronments could be used just as easily in scaled-out arrangements, where
workloads aren’t at all mixed. The reduction in per-user costs doesn’t, at
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 55
the moment, come from using different computer equipment or different
operating systems: It comes from running a small number (or even just one)
workload and scaling it up as much as possible. That’s how cloud computing
reduces costs dramatically.
No corporation that runs a mixed workload is ever going to achieve cloud
computing’s economies of scale.
But how do massively scaled data centers manage to get their per-user costs
so very low? This becomes clear when you read about each area of data
costs in Chapter 21.
Comparing Traditional and
Cloud Data Center Costs
Before reading how to reduce data center costs, reread the traditional IT
costs statistics:
✓ Portion of IT budget used to maintain and run existing systems:
70–80 percent
✓ Portion of IT budget used to build and implement new capabilities:
20–30 percent
Compare traditional and cloud data centers in Table 5-1.
Table 5-1 A Comparison of Corporate and Cloud Data Centers
Traditional Corporate Data Center Cloud Data Center
Thousands of different Few applications (maybe even just one)
applications
Mixed hardware environment Homogeneous hardware environment
Multiple management tools Standardized management tools
Frequent application patching and Minimal application patching and updating
updating
Complex workloads Simple workloads
Multiple software architectures Single standard software architecture
56 Part II: Understanding the Nature of the Cloud
Looking at the table, it becomes clear that the cloud data center is much sim-
pler to organize and operate and, because it is simple, it scales well. In other
words, the larger you make it, the lower the costs per user are. In the next
section, we examine some of these costs and see where the efficiencies arise.
Examining labor costs and productivity
Labor costs depend on several things:
✓ Technology managing the data center: Even improving that technology
in a traditional corporate setting may reduce the cost of labor only a
small amount.
✓ In what environment someone works: The labor cost per person is
likely to be equivalent regardless of the data center type; the skills
requirement is the same. But that person’s productivity varies depend-
ing on the environment. Operating the scaled cloud data center is much
simpler.
The impact of this set of differences on labor costs is dramatic. Corporate
data centers usually have a ratio of operational staff to severs of around 1
person to 65 servers. In cloud data centers, that ratio is more like 1 person
to 850 servers, and we’ve even come across better ratios than that. This is a
10-to-1 improvement in the productivity of labor (or possibly more — maybe
going as high as 20 to 1).
Wondering where you are
The traditional setup’s 58 percent costs depend a lot on location:
✓ Electricity fees
✓ Local taxes
✓ Labor costs
Compare a data center in North Carolina with one in New York (keeping in
mind no two data centers have the same software workloads). Better to con-
sider technology costs separately and see where economies arise, which we
do in the following sections.
Chapter 5: Seeing the Advantages of the Highly Scaled Data Center 57
Electric power
Computers have been using more electricity in recent years and, at 7 percent
of corporate data centers’ costs (including heating and cooling), the cost is
significant. Cloud data centers use even more: Electricity costs hover around
12 percent.
Cloud data centers can do the following:
✓ Put the data center where the cheap power is. Electricity fluctuates in
price from year to year and costs are difficult to control.
✓ Negotiate a discounted power contract with its power company.
Cloud data centers, by their level of usage, fall into the least expensive
category.
If a cloud data center is contemplating building a data center,it can
negotiate a long-term deal for an even deeper discount than industrial
usage gives them. Put the data center very close to the power plant and
bargain for a lower cost supply based on these points:
• Distance from the power station (because less electrical power is
lost in transit).
• Minimal power interruption from electrical storms (if you have a
private circuit direct to the power station).
Outsourcing
Because power is so critical to the cloud data ✓ Oil prices change, which can cause cost
center, organizations have to consider the fluctuation.
availability and cost of energy sources as they
✓ Liquified natural gas (LNG) suffers from the
would any primary data center resource.
same changing fuel prices as oil.
Electricity sources include the following:
✓ Coal is more stable in price, but not green.
✓ Hydroelectric is generally expensive when
✓ Nuclear is inexpensive to run but expensive
it has to travel far to customers, but other-
to build and gain approval
wise it’s usually cheap and can be the ideal
source of power for a data center.
58 Part II: Understanding the Nature of the Cloud
Other location costs
Other location related costs when building a new data center include the
following:
✓ Land costs: The days of siting data centers in skyscrapers in Manhattan
are over. Better to use cheap land with low property taxes. There are
exceptions, of course. For example, in algorithmic financial trading,
latency lost due to networking (communications) distance directly
impacts revenue.
✓ Building costs: A designed-entirely-as-a-data-center building is a must.
• Heat management is the overriding priority, so building out almost
certainly makes more sense than building up. Cool geographical
areas may make more sense than hotter ones.
• Safety is another important consideration. Data centers need to be
electrically safe, secure, and fireproof.
✓ Staff: Although staff costs are very low for the cloud data center, as a
percentage of the whole, location in areas (or even countries) where
staff costs are low can further reduce staff costs.
✓ Investment incentives and taxation: Many areas of the world, including
states in the United States, welcome inward investment and help finance
it with very generous tax exemptions and cash incentives. Take advan-
tage of these opportunities when you find them.
In the next chapter, we examine technology costs, which also favor the cloud
data center in many ways. The simple fact is that data centers as they exist
now, in the enterprise, are a cottage industry that’s going to change in the
coming years by the mass-production efficiencies of cloud data centers.
Chapter 6
Exploring the Technical
Foundation for Scaling
Computer Systems
In This Chapter
▶ Comparing traditional data centers to clouds
▶ Achieving economies of scale
▶ Saving money via the bottom line
I n Chapter 5, we contrast the non-technology operational costs of the tra-
ditional data center with those of the cloud data center (electricity, cool-
ing, space, and so on). In this chapter, we contrast technology costs between
the traditional data center and the cloud data center.
We divided into four areas the places where IT spends money:
✓ Hardware, including servers, storage, and so on
✓ A power supply for those systems and how to keep them from overheating
✓ Networking and communications equipment so the systems can
interoperate
✓ Electricity to support the overall data center
Some elements are more expensive than others. In Chapter 5, we look at two
reports that detail the costs of running traditional and cloud data centers.
Using this same set of numbers, we calculated the costs of the areas. The
results are quite interesting. The greatest expense in the traditional data
center is server and storage hardware, which accounts for 36 percent of the
amortized costs. The second biggest expense? Power distribution and cool-
ing. Amortized over a year, power and cooling are 20 percent of the total
60 Part II: Understanding the Nature of the Cloud
expenses. Both networking and electricity each add 12 percent to the total
expense number per year. Add hardware and its supporting power and cool-
ing, and you have 56 percent of the technology related costs.
We discuss electricity costs in the previous chapter, but only from the per-
spective of arranging for an inexpensive supply. In this chapter, we take on
the issue of using that electricity efficiently.
Server-ing Up Some Hardware
Although we’d like to tell you that costs are static, clearly they aren’t. Costs
for your data center hardware will vary dramatically depending on the type of
workloads you support.
Data storage is an excellent example of this variation. If a data center is feed-
ing video to the Internet from a vast video library (like YouTube does) the
storage requirements are huge. However, storing short text messages (as
Twitter does) doesn’t require a lot of space. Indeed, Twitter doesn’t even
store its billions of messages indefinitely. The YouTube library, on the other
hand, just keeps on growing.
Tradition! versus clouds
What does this mean when you look at the differences in the costs of hard-
ware between the traditional data center and the cloud data center? Look at
a snapshot of each:
✓ Tradition: In a traditional data center, IT management has a structured
process for purchasing hardware. Each year they talk to business units,
determine what new applications to add or expand, and work with ven-
dors on the procurement process. In addition, most IT organizations
refresh their hardware on a regular basis to make sure that things run
smoothly and old systems are retired before they cause problems.
✓ Cloud: When a business is creating a cloud data center (either a private
one inside the firewall or a service provider) the process of procuring
systems is very different. Because the cloud supports very different
workloads, IT management doesn’t buy traditional hardware. Rather, IT
management might go directly to an engineering company that designs
the system boards and networking switches for them, and then take the
contract to a manufacturer to have them build the precise hardware
that they want.
Chapter 6: Exploring the Technical Foundation for Scaling Computer Systems 61
The bottom line is that the cloud data center is well suited to buying precisely
what you need in a very economical manner. In contrast, the traditional data
center doesn’t have the same economies of scale.
We aren’t being critical of the server products that are built and delivered by
big computer manufacturers. Such engineering is difficult to criticize in its nat-
ural context. All such servers, whether mainframes or cheap commodity
server boards, are designed for general circumstances of typical customers.
It’s just really unlikely that the requirements of a cloud center are anywhere
close to typical.
Considering cloud hardware
When your company is establishing a cloud data center, think about the
hardware elements in a different way. The following sections summarize
considerations.
Cooling
Cloud data centers have the luxury of being able to engineer the way systems
(boards, chips, and more) are cooled. When systems are cooled via air condi-
tioning, they require tremendous amounts of power. However, purpose-built
cloud data centers can be engineered to be cooled by water, for example
(which is 3,000 times more efficient than air in cooling equipment).
CPU, memory, and local disk
Traditional data tends to be filled with a lot of surplus equipment (either
to support unanticipated workloads or because an application or process
wasn’t engineered to be efficient). Surplus memory, CPUs, and disks take up
valuable space and, of course, they need to be cooled. The cloud data center
typically supports self-service provisioning of resources so capacity is added
only when you need it.
Data storage and networking
Data storage and networking need to be managed collectively if they’re going
to be efficient. This problem has complicated the way the traditional data
centers have been managed, and has forced organizations to buy a lot of
additional hardware and software. The cloud data center can be engineered
to overcome this problem. The cloud knows where its data needs to be
because it is so efficient in the way it manages workloads. The cloud actually
is engineered to manage data efficiently.
62 Part II: Understanding the Nature of the Cloud
Redundancy
Data centers must always move data around the network for backup and
disaster recovery. Traditional data centers support so many different work-
loads that many approaches to backup and recovery have to be taken. This
makes backing up and recovering data complicated and expensive. The
cloud, in contrast, is designed to handle data workloads consistently. For
example, in a cloud data center you can establish a global policy about how
and when backups will be handled. This can be then handled in an automated
manner, reducing the cost of handling backup and recovery.
Software embedded within the data center
We talk a lot about software in the context of applications, but a considerable
amount of software is linked at a systems level. This type of system level soft-
ware is a big cost in the traditional data center simply because there are so
many more workloads with so many operating systems and related software
elements.
As you know, cloud data centers have fewer elements because they have
simpler workloads. There are some differences in how software costs are
managed depending on the type of cloud model. Cloud providers understand
these costs well and design their offerings to maximize revenue. It will help
you understand pricing by understanding the cost factors for each of the
models.
The following gives you a sense of the difference between IaaS, PaaS, and
SaaS when it comes to embedded software costs:
✓ An Infrastructure as a Service (IaaS) operation is likely to have higher
software costs because although it provides only an environment for
running applications, it has to build that environment according to
equivalent environments in corporate data centers. Therefore, the IaaS
vendor has to spend a lot of resources on management and security
software in addition to the operating systems. See Chapter 10 for more
about IaaS.
✓ With a Platform as a Service (PaaS) operation, the provider delivers a
full software stack. To reduce cost, the PaaS vendor is likely to provide
a software stack consisting of proprietary components. The licens-
ing costs may be lower for IaaS than the PaaS environment because
the operator is likely to force the use of specific software products.
However, the PaaS vendor must maintain and support the software stack
it provides. See Chapter 11 for more about PaaS.
Chapter 6: Exploring the Technical Foundation for Scaling Computer Systems 63
✓ With Software as a Service (SaaS), the SaaS vendor provides a propri-
etary application as its value to customers. While the vendor invests in
this software, it typically relies on partners to support many of the other
functions. These vendors also take advantage of open-source compo-
nents. See Chapter 12 for more about SaaS.
Open-source dynamic
The cloud is an economic and business model as much as a technology model.
It isn’t surprising, then, that open-source software is an important element
for almost all cloud providers. Some of it is very high quality and nearly all of
it can be used for no license fee, as long as you obey the restrictions of the
associated license.
Open-source software has already become a business factor in the Internet
service provider (ISP) business, with most ISPs providing an easily installed,
highly functional software stack for building Web sites. Many cloud providers
take open-source software as a foundation and customize it to optimize sup-
port for their workloads.
The other software area that impacts costs is the way operating systems
are handled in the data center. Under traditional operation, an OS has many
background processes running. All such processes have a function and quite
a few of them run by default, whether you need them or not. Some of them
are keeping logs, some are handling messages from the network, some fire off
scheduled jobs, some handle printing, some provide directory services, and
so on. They all sit there happily chewing up CPU cycles. None of them should
be there unless they have a specific role to play.
In a traditional environment, no one would think of deleting useful background
processes, but nothing superfluous should run in an environment that prizes
efficient resource usage. Not only that, but if you’re running a cloud data
center, you may be interested in rewriting some of these tasks because you
need them to run slightly differently. That’s why open source plays a large
role in cloud operations.
Economies of Scale
We spend a lot of time in this chapter saying why the economics of the cloud
are so different than that of the traditional data center. Of course, not every
workload is right for the cloud.