Virtualization: Transforming the IT Landscape

By Diane Greene
President, VMware

Virtualization: Transforming the IT Landscape

Industry-Wide IT Adoption
The 1990s set the stage for the current demand for
virtualization software. Contributing factors included the
availability of standardized, high-volume, and increasingly fast
Intel and AMD processors, ubiquitous Microsoft and Linux
operating systems, low cost storage and memory, full
distributed systems support, Ethernet, NFS and the World Wide
Web.
VMware revisited the virtual machines that IBM pioneered for
mainframe systems in the late 1960s and early 1970s. Since the
x86 architecture did not support virtualization and using virtual
machines on mainframes required a proprietary operating
system and hardware, the technology did not achieve market
success.
VMware changed this by enabling virtualization on industry
standard components without requiring changes to industrystandard processors or operating systems. VMware invented the
hosted model for desktop virtualization software with VMware
Workstation. The company added bare metal virtualization for
servers in 2001 with ESX Server and invented VMotion for the
live migration of virtual machines. VMware also
introduced virtual appliancespre-installed and pre-configured
application and operating system environments that run on
virtual machinesthat are distributed in a similar fashion to
the popular software as a service (SaaS) distribution model.
VMwares newest offering, VMware Infrastructure 3, provides a
true distributed virtualization platform for the data center.

Modern Virtualization in Two Phases

VMware recently entered a second phase for advancing
virtualization software and technology. The first phase started in
1998. During this period the company focused on
evangelizing the benefits of virtualization and delivering highly
robust and functional products to the marketplace. More
than eight years later, virtualization is now one of the hottest
trends in the IT industry. People trust virtualization and are well
acquainted with the benefitsincluding cost savings, simplified
resources and the freedom to choose the hardware or software
that users want to operate. VMware no longer needs to
evangelize the value of virtualization. For the second phase,
VMware is articulating the value of its products and setting the
agenda for the future of virtualization. The company
continues to deliver phenomenal advances to our
customers. Our products transform how VMware customers run
their systems and manage their softwareproviding ROI within
3-6 months of their virtualization investment.

ESX Server
In 2001, with the release of ESX Server, VMware introduced a
single node hypervisor. ESX Server allows users to partition a
physical machine and run multiple operating system
application combinations within a virtual environment. ESX
Server enables users to increase the CPU utilization of their
server hardware from 5-15 percent to up to 60-85 percent.
ESX Server also lets IT departments provision new servers in a
matter of minutes instead of days.

Virtual Machines
VMware virtual machines run on any hardware configuration,
whether a 2-way CPU, a 32-way CPU, a blade system or a laptop.
This removes much of the logistical overhead burden when
implementing system-wide initiatives such as disaster recovery.

Provisioning
With the advent of VMware virtual infrastructure users can
provision virtual machines onto a single software layer that can
span different hardware components.

VMotion
With VMotion, users can take a running virtual machine and
move it without service interruption across physical machines.
For example, a company can be serving a movie and the clients
receiving the feed will experience no loss of picture quality
as the movie server is moved from one machine to another.
VMotion also allows IT departments to service a machine in the
middle of the day without client interruption.
VMware virtual infrastructure and VMotion offer capabilities far
beyond the single node hypervisor.

Virtualization Has Become Mainstream

Today, more than one million servers are virtualized into
VMware virtual machines.
Those systems are highly stable. For example, VMware has a
large pharmaceutical customer that has been running ESX
Server continuously with no reboots or faults for over 800 days.
VMware is also doing its part to sustain the planet. The
company estimates that VMware virtual machines cumulatively
have saved over 5 billion Kilowatt hours of power. As cooling
and space become mission critical issues for the data center,
the consolidation and power saving capabilities of VMware
software are increasingly important.

Server, Storage and Network Consolidation

In 2003, the IT service department of a leading North American
utility company started a server consolidation project using
VMware software. More than 1,000 servers were migrated to 50,
8-socket capable IBM x440 servers.
Virtualization-based consolidation not only affects servers,
it also raises the quality and service levels of the entire data
centerincluding storage, network and facilities.
For the utility company example, storage was consolidated
from silos of internal disks to a highly available tiered storage
solutiona Hitachi SAN. There was also a 10:1 consolidation of
network ports. The customer experienced dramatic cost savings
in their facilities and hardware infrastructure.

Impact
The net impact of this customer example includes over $8M
in quantifiable cost savings. While data center power
savings is a popular topicwhat goes unnoticed is the
tremendous cost avoidance virtualization brings to power.
At the same time, customers deploying virtualization
solutions are starting to realize that virtualization is not an
ROI-based one-time consolidation/cost avoidance project,
but rather an ongoing strategy for operational efficiency.
For instance, in just the hardware refresh cycle alone, the
customer in this example saves 10 thousand man hours per
year, allowing resources to be reallocated to more strategic
tasks.
VMware customers often overlook the incredible resource
utilization capabilities and cost savings. They are now
looking at their virtualization stack as a means to simplify IT
resources and make their processes more efficient.
While hard cost savings are important and proven in
virtualization deployments, VMware customers are
increasingly asking for the efficiency benefits of
virtualization. IDC says labor, and in most IT organization
resources, surpass capital spending by more than 3:1.
Meanwhile, management and administration labors costs
are growing at 10 percent year over year verses 3 percent
for industry standard infrastructure.

Transforming Hardware and Capacity

Mangement
This example shows traditional servers and static provisioning.
People will look back in a few years on these systems and the
methods for maintaining them as old fashioned and the old
way.
On physical servers, an application is carefully and statically
provisioned onto a roughly right hardware box. The resources
typically end up siloed with over-provisioned capacity that was
defined by the chosen hardware. Capacity planning and
estimation in most cases is not done, at all. If users need to
re-provision or upgrade capacity these resources are inflexible
and require application downtime and manual hardware/
application migration.
The earlier generation of virtualization software offered a much
better situation for provisioning resources, with hardware
provisioned separately from the application. The robust single
node resource management capabilities built into VMware ESX
Server allows applications to dynamically and fairly share
capacity of modern generation industry standard hardware.
Using VMware VMotion and networked storage, users can
manually re-provision or migrate a workload to new servers
without any downtime.

This eliminates the problem of over-provisioning and gives

users the operational flexibility of non-disruptive re-provisioning
of a workload, leading to the dramatic cost saving discussed
earlier.
While the boundaries of hardware resources have started
to blur there is still room for improvement. In maintaining a
virtualized environment, IT administrators still need to think
about which physical hardware a virtual workload resides, or on
which physical server the workload needs to be placed. When
rebalancing workloads, administrators need to be aware of the
physical nature of the system (is it an old CPU, a 2-way system
or 8-way system) as well as the status of the existing workloads
on given server hardware.
Now, with VMware Infrastructure 3, IT organizations can enable
a true utility model for delivering compute capacity for multiple
application groups and businesses. IT labor around capacity
management is dramatically simplified, and IT organizations
and their customers can eliminate the problems of maintaining
physical hardware.
VMware Infrastructure 3 introduces two ground breaking
technologies that take the labor and thought out of
capacity management, enabling a true utility model for
managing shared hardware. VMware Distributed Resource
Scheduler (DRS) intelligently allocates resources for virtual
machines across a farm of physical servers. A Logical Resource
Pool is a facility that defines logical, flexible resource and service
level policies, and allows VMware DRS to distribute capacity to
varying business needs.
Administrators and their customers no longer need to worry
about the underlying physical nature or configuration of the
server farm.

Logical Resource Pools

From the physical resource pool accessed by VMware
Infrastructure 3, users can create any number of logical resource
pools. IT organizations and their customers can specify a logical
sub-quantity of hardware capacity that is then intelligently and
automatically mapped into the physical resource pool.
IT organizations can aggregate a collection of industry standard
systems and treat them as one large computer that can then
flexibly meet varying business needs and service levels.

VMware DRS
VMware DRS dynamically allocates and balances computing
capacity across a physical resource pool.
VMware DRS continuously monitors utilization across resource
pools and intelligently allocates available resources among
the virtual machines based on pre-defined rules that reflect
business needs and changing priorities.
IT managers no longer need to think about where capacity
resides or on which physical server a workload is running. Their
only concern is the quantity of compute capacity that a given
business group or set of applications require.

Automate Resource Assurance for Critical

Applications
With VMware DRS, hardware becomes a truly distributed,
pooled resource where applications get all of the computing
cycles they need to execute with peak performance. IT
organizations can run mixed workloads on industry standard
systems and guarantee that the highest priority applications
will get the resources needed for optimal performance. CPU
capacity, disk size, network bandwidth, and memory capacity
are assigned to each virtual machine in accordance with its
peak load requirements. This allocation can occur for a group
of applications, as well. If the applications are not using peak
load, they will share the resources with less important task
operations.

For example, if an application is mission critical and has been

allocated nearly all of the CPU cycles and memory available
from the first server, as it starts to use all available resources,
the system will automatically take the other virtual machines
on that system and move them to another system that has the
headroom to support them. The highest priority application
will be guaranteed to receive the resources it needs. Note: this
is not policy-based. Meaning, it does not use a pre-defined
behavioral constraint as means of management control.
However, it will allow users to determine the resources required
by each application and guarantee that those resources are
available.

Non Disruptive Capacity on Demand

Using VMware DRS, potential problems are diagnosed and
corrected as soon as they occurwhen an application
experiences increased load, and before it experiences a serious
performance issue, VMware DRS detects the event, and
provides more resources to the virtual machine. This capability
enables a self-adjusting, self optimizing data center.
System administrators can allocate available resources
depending on the value of the applications they support. Not
all applications are created equalsome are more important
than the others. With VMware DRS administrators can ensure
that the important applications always take precedence if there
is competition for resources.

10

Automatic Availability for all Applications

VMware has added built-in cluster management to VMware
Infrastructure 3. This means that any application, by virtue of
operating inside a VMware virtual machine, will have automatic
restart. If the hardware fails, the virtual machines will be
automatically restarted on a server that has the headroom to
handle the application load of those virtual machines.

Transform Availability Service Levels

Traditional cluster management is a heavy overhead
configuration process with an application specific configuration.
It does handle state so that at restart, no state is lost. However,
with typical cluster management less than 10 percent of
applications have fault handling capabilities.
With VMware cluster management, a failure will cause the
application to be restarted on a running machine so that every
application will have application availability through automatic
restart. This increases the level of availability without adding any
complexity to the system and without requiring administrators
to have special application knowledge.

11

Backup Anytime
VMware Infrastructure 3 also includes an agentless backup
where users do not need to install a separate agent per virtual
machine. The backup feature uses VMware virtual machine
snapshot technology to backup from a host while the
application is online.
Backup can be conducted at any time, because it is completely
separate from production virtual machines, with no agents
running in the virtual machines to hinder network traffic or clog
the LAN. The backup process is moved off the virtual machine
and the network onto a separate proxy server that does not
interfere with the performance of the production systems.
This reduces the load on ESX Server installations, allowing them
to run more efficiently and run more virtual machines.

12

Fault Tolerant Data Center

VMware Infrastructure 3 offers a true, distributed virtualized
data center and it is based on industry commodity parts. If
a part fails, the only consequence is a few less resources are
available for the time that the part is down. Other than that, the
fault is transparent and everything continues to run. Systems
can be added or removed and workloads can be increased or
decreased without requiring any manual reconfiguration of the
system.
All physical resources are managed separately from how
resources get allocated to individual workloads or applications.
Scaling up or down in storage, network and server resources
happens independently of how those resources are mapped
to individual applications. Workloads and applications simply
consume their allocated percentage of physical resources. If
a resource pool suddenly shrinks, they will continue to run,
but they will only use a fraction of the previously available
resources. If the resource pool grows, they can dynamically
grow their resource consumption to take advantage of the
extra capacity.
As a result, the resources of an enterprise can be efficiently
managed to meet application service level agreements
without excessive downtime, maintenance windows or
change management. If an Exchange Server is struggling, for
example, and IT decides to scale its resource allocation, it can
get additional computing, networking and storage resources
without any form of downtime. If IT wants to take resources
away from one application to support another, they can do
so, on the fly and without incurring downtime on either of the
applications. It is, simply, a far better way to manage data center
resources.

13

Extend Virtual Infrastructure to End-User

Desktops
Another important area for customers is securing and
centralizing enterprise desktop management. Customers
have been using VMware virtual infrastructure to host virtual
machine desktops from a server for several years now and many
customers host thousands of virtual machines. This capability
is particularly useful for organizations with remote workers,
temporary workers, and test and development teams. Since
the virtual machine runs a full instance of unmodified Windows
or Linux, every application runs as it would on real hardware.
Recently companies have been evaluating virtual enterprise
desktops for emergency situations (for example, a natural
disaster or flu epidemic) when employees would be required to
work from home.
With the added capabilities of VMware Infrastructure 3, hosted
desktops will get automatic load balancing and high availability
plus scalability and performance. VMware has also added a
per-client licensing framework for hosted desktops.

Application Lifecycle Management

Customers have been using VMware products for over seven
years to accelerate test and development, decrease the burden
of configuring for possible test scenarios and increase their
ability to support the growing number of possible operating
system combinations. Many VMware Infrastructure customers
are using our products to streamline their entire software life
cycle. These customers take the developed software and test
it in a virtual machine so no differences are introduced in the
installation and configuration process. The same virtual machine
can go to staging and production. Production can be rolled
back if necessary and also incorporate the never-ending cycle
of updates. Software companies are also using virtual machines
to pre-install and configure their software for delivery to
customers. It is a full software life cycle that is streamlined and
made consistent through the use of virtual machines.
Recently VMware announced the acquisition of Akimbi, a small
company devoted to building tools that streamline the virtual
machine software life cycle. VMware will continue to invest in
this area to expand the significant software life cycle benefits
that VMware products and technology already make possible.

14

Virtual Applications Changing the Role of the

OS
The traditional software stack is comprised of hardware, an
operating system and an application. Traditionally, the
operating system must manage both the hardware and the
application. This is a highly complex job. VMware virtualization
is a tremendous simplifying agent here. If we personify the
software stack it might look like this: One person takes new
hardware from the loading dock and gets it installed. Another
person might get the operating systems installed and
configured and there someone else architects the application
stack to run on top. It would be a highly complex job for one
person to do all of these things. In effect, this is what VMware
does with adding the virtualization layer.

The hypervisor takes care of the hardware and the operating

system takes care of the application.

Application designers choose the operating system that is

optimal for their application. Say they have a highly secure
application, they will chose a highly secure operating system.
Alternately, if designers have a high workload, then they will use
a very low overhead, bare bones operating system. Instead of
a one- size-fits-all operating system, the industry will see many
operating systems tailored to the applications that they run.
This will also result in more stable operating systems as they
will be built for specific application loads. In this scenario, virtual
appliances offer an ideal solution and a simpler model. The
operating system bundle can be contained in a virtual
appliance, running on the virtual infrastructure and underneath,
the hardware.

15

The Distribution of Virtual Appliances

Users will group their virtual appliances in a Web service or in
multiple instances of email. Virtual appliances will be
provisioned and resources for them will be allocated in a similar
fashion.

VMware has set up a portal for virtual appliances and we now

have hundreds available. Appliances are available for safe Web
browsing, network firewalls, IP telephones, fax applications and
more. VMware launched the Ultimate Virtual Appliance Contest
a few months ago with $200,000 in prize money. VMware has
hundreds of submissions and a highly esteemed outside panel
of judges will select the winner, to be announced at Linux
World in August.

16

The Value of VMware Infrastructure

In summary, traditionally there are servers, storage arrays,
routers, and lots of cables all carefully and statically hooked
together. Applications are provisioned to a singe server with its
available, attached other resources. Once they are running, they
are left alone with the hope that the workload does not change
and the systems remain stable.

With VMware infrastructure everything becomes hardware.

VMware infrastructure sits on top of all that hardware and

exports it to the virtual appliances as a single pool resources
100MHZ of cycles, 100GBytes of memory, etc.

17

This simple model takes great complexity out of the data center
while also adding significant utilization optimization and better
data continuity through more ubiquitous availability and online
backup.

VMware builds robust and highly functional products.

We are first and foremost a partnering company. We also will
continue to invent and develop with new ways to leverage our
fundamental virtualization building blocks.
VMware is interested in seeing an open industry. The company
views this as a unique moment in time where we can have
an industry where customers choose solutions based on
functionality, quality, and price. They are not forced into a
solution through arbitrary lock-in or control points.
To that end, VMware is investing heavily and proactively
working toward standards that are license free and governed
by an independent industry standards body. We have the
defacto standard, over 1M servers virtualized in our virtual
machines. But we also feel very well equipped to compete
where there is freedom of choice and no lock-in.

Virtualization Standards
There are three standards that are relevant right now to the
virtualization industry.
First, is the interface between the operating system and the
hypervisor.

18

When VMware started out there was no possibility for altering

an operating system to work better with a hypervisor. Our first
prototype worked with a modified Linux kernel but we knew
that we had to have a product that would work with existing
Linux ,Windows, Solaris, Novell--and now Mac
operating systems. No matter what, we want virtualization
software that runs legacy workloads without requiring any
porting or modifications.
Now that virtualization is accepted, operating system vendors
are also building hypervisors and they are modifying the
operating systems to run more optimally. The interface should
be open, license free, and governed by a third-party standards
body. Any operating system should run optimally on any
hypervisor without arbitrary licensing restrictions.

The next interface is around the management of the virtual

machines. The management software needs to be able to
communicate with the hypervisor to manage and manipulate
the virtual machines. This too should be governed by open and
third party standards body. VMware has offered its interfaces
and APIsbased on more than seven years of production
product experienceto be available and governed by an
industry-neutral body. We have done the same around the
hypervisor/OS interface.

Finally, the third standard is for the format of the virtual

machine itself.

19

It does not particularly matter what the interface is but the

standards will be used to determine how a virtual machine
is backed up, patched, recovered, etc. The format should not
require a license as it would be an arbitrary control point.
The virtual machine format should likewise be governed by a
neutral standards body.

VMware is also working to establish an industry-wide,

standardized set of benchmarks. Our goal is to establish
benchmarking standards that give meaningful workload
information so customers and partners can make clear applesto-apples comparisons between different virtualization
solutions.

20

VMware Vision
VMware virtualization has moved way beyond the single node
hypervisor. With VMware virtualization, users can wrap an entire
service-oriented application with all its moving parts into a
virtual appliance. This appliance can be directly placed in a
resource pool. The virtual appliance is equipped with built-in
provisioning, resource management, availability, security and
data protection features. The dial on all these capabilities can
be turned up or down at the click of a mouse.

21

VMware, Inc. 3145 Porter Drive Palo Alto CA 94304 USA Tel 650-475-5000 Fax 650-475-5001 www.vmware.com
2006 VMware, Inc. All rights reserved. Protected by one or more of U.S. Patent Nos. 6,397,242, 6,496,847, 6,704,925,
6,711,672, 6,725,289, 6,735,601, 6,785,886, 6,789,156 and 6,795,966; patents pending. VMware, the VMware boxes logo
and design, Virtual SMP and VMotion are registered trademarks or trademarks of VMware, Inc. in the United States and/or
other jurisdictions. Microsoft, Windows and Windows NT are registered trademarks of Microsoft Corporation. Linux is a
registered trademark of Linus Torvalds. All other marks and names mentioned herein may be trademarks of their
respective companies.