Exchange Virtualization Methodology
Successfully deploying Exchange on vSphere using proven methods
Author: Brandon Noeldner
Introduction vSphere enabled the leveraging of Windows 2008 ability
Server virtualization has become a standard for many organizations to dynamically scale through addition of key components
to reduce capital and operational costs, increase corporate agility, (Compute, Communication and Storage) without
and increase service availability. Early in the adoption of interruption.
virtualization technology, efforts are directed towards the Decreased complexity through standardized building
consolidation of underutilized servers because they addressed the blocks for highly available environments.
most immediate business need of efficient resource use. As New hardware can easily be integrated into the
virtualization technology penetrates the data center further, the environment without extensive testing; enabling future
drivers of adoption become less about cost, and more about performance benefits to be leveraged with little effort.
improvement of availability and performance. Messaging is often Recoverability
not considered as a potential candidate due to its intense By making the application hardware agnostic,
availability and performance requirements. To counter these claims, recoverability is increased by simplifying the requirements
Data Strategy leverages a proven methodology to not only reduce for full system restores.
overall costs, but increase availability, and improve performance. Testing is greatly simplified through leveraging key
When combined with the performance enhancements made in components of VMware Infrastructure such as SRM and
vSphere 4 and Microsoft Exchange 2010, combined with changes in automated testing.
processor technology, the argument to virtualize becomes even
more compelling. Exchange is a good example of a tier 1
application that benefits from being virtualized on a properly Methodology
When looking into the virtualization of Exchange, the load
Justification characteristics must be well understood. A comprehensive
The primary benefits of virtualizing Exchange can be summarized in understanding of the number of mailboxes and I/O requirements
three categories, which vary in importance based on the current should be held by those implementing any Exchange Migration.
business needs. Proper use case generation is essential to business critical
Performance applications such as messaging. In cases, adequate compute
Virtualization enables higher utilization of physical server resources may be starved for storage IO (2). To avoid such issues, it
compute resources by enabling an optimal distribution of
is important to address the core needs of the application through
resources. Performance testing has shown a single ESX
host running is capable of running up to 16,000 mailboxes proper assessment of the needed environment. While the
in LoadGen tests, while maintaining sub 600ms Sendmail methodology used is the same, the value drivers will differ with a
response times.(1) tier 1 application. The methodology can be broken down into the
Availability following steps.
Virtualization enables an environment to reduce planned
down time and the effect of introducing new hardware Assessment
since the application becomes hardware agnostic. Information gathering stage. Includes collecting
Manageability performance metrics and possible interview process with
Virtualization simplifies future hardware migration by affected parties.
making applications hardware agnostic. New hardware Designing the Solution
may be leveraged with minimal effect on the overall Cultivating the requirements gathered from the
Exchange configuration. Assessment to properly size the environment, and design
Streamline testing and troubleshooting by utilizing to include any business requirements.
snapshots and clones to test updates and patches. Test Implementation
environments can be readily created by cloning the Building out servers per the design, and following a user
existing production environment. migration plan
vCenter decreases deployment time through leveraging
templates to shorten provisioning times to minutes, and
reducing the variance in deployment.
Assessment usage patterns are understood, a combination of perfmon along
with scheduled tasks will allow you to capture data points over a
The Assessment stage involves building an understanding of the wide period of time.
businesses expectations of their messaging environment In order to
properly size and scope an environment one should understand the
- How many mailboxes / exchange users Designing the Solution
- Approximately how many messages are sent and received The plan and design stage is where you will build out a blueprint
in a day per user and process for deploying the application using vSphere. Take into
- Maximum mailbox size allowed for users consideration the information gathered throughout the assessment
stage to properly size the virtual machines, and implement any
Once the required performance is met, the required warranty changes that would be needed to the vSphere environment to
should be addressed concerning follow best practices.
- Compliance issues.
- Availability requirement. Design Options
Care must be taken to properly understand the load requirements Depending on the business requirements, Exchange paired with
of Exchange to ensure the environment is adequately sized. During vSphere offer several options for availability. There is a tradition
the assessment, performance metrics are collected from the mail server clustering supported by Exchange 2007 (SCC, CCR, LCR,
existing physical environment, and interviews of the interested SCR) or in Exchange 2010 (DAG). New in vSphere is the ability to
parties are performed so that both technical and business use Fault Tolerance to provide another availability option without
requirements are understood. the complexity of enabling Microsoft clustering, and with an almost
instant failover window.
Understand Business Expectations
It is important to understand the business drivers and expectations Server Sizing
when implementing a messaging solution of this scale. It is possible Based on the information provided from the assessment, proper
that a deployed solution that meets or exceeds the performance server sizes can be determined. VMware’s best practices suggest
metrics gathered is deemed a failure, because business using a standard building block size when building Exchange virtual
expectations were not fully understood, and as a result not met. servers. This allows for right sizing your servers rather than having
wasted resources from over provisioned physical Exchange servers.
By having a pre-defined virtual machine with a known mailbox
The Operational Level Agreement (OLA) requirements, along with
capacity (500, 1000, 2000, etc) one is able to architect a modular
understanding the business needs in regards to RPO and RTO, and and flexible design that scales easily, and delivers predictable
how they are tied to Exchange must be understood to ensure the performance when growing to meet evolving demands in the
end result matches expectations. Recovery Point Objective (RPO) is future. Microsoft offers guidance on how to size server blocks based
defined as the maximum acceptable level of data loss following an on the number of users, type of users, and which roles the servers
unplanned event. The point in time snapshot before a disaster that will host (3). With the performance information collected from
Exchange, it must be determined if the Virtual Infrastructure has
Exchange can be restored will be the recovery point. Recovery Time
enough capacity in its current configuration, or if modifications are
Objective (RTO) the period of time between disaster, and normal
required to meet the higher demands of Exchange.
operations are resumed. The number of emails that are lost in
between this point in time and the event will vary on the backup Always remember to follow the support guidelines that Microsoft
architecture, but must be understood so that the end design meets has set when introducing Exchange into a virtual environment.
business expectations. Building out a supported configuration not only ensures that you
will have the best performance when moving into production, but
will also be in full compliance in the event that a support case needs
to be opened.
Without knowing the performance characteristics of the current
Exchange servers, there is no way of measuring the improvements
of moving to vSphere. A performance profile can be built by
Building out a blueprint for the new Exchange configuration will act
utilizing performance monitor to examine CPU usage, Memory
as a roadmap, and visual representation of how the application will
utilization, and Disk IO stats, It is possible that the physical
fit into the virtual environment. Going forward, proper
hardware that Exchange is currently running on is oversized, and if
documentation can be used to help troubleshoot and understand
the physical sizing was emulated when building out Exchange on
the environment in the future. This is also where you can document
vSphere, un-needed waste may occur.
any component of Exchange that is not being virtualized, and give
justifications for why it could not be virtualized. An example of this
Performance counters will not tell the whole story however if
would be a hardware network load balancer that is being used to
samples do not incorporate all use cases. It is possible that data is
balance requests to a cluster of Client Access servers.
collected over a period of time that does not see peak usage. It is
important to understand the usage cases before interpreting
Part of the documentation process includes building a migration
performance results. Knowing when peak demand is seen such as
methodology from the old system to the new virtual Exchange
end of month / quarter, or during product roll-outs will better help
install. The downtime involved here must not break OLA’s and
plan for worst case scenarios for sizing Exchange servers. Once the