CASE STUDY 7 DATA CENTER CONSOLIDATION AT GUARDIAN LIFE
As one of the largest mutual life insurance firms in the United States,
Guardian Life (www.guardianlife.com) has more than 5000 employees and
over 3000 financial representatives in 80 agencies. Guardian and its
subsidiaries provide almost three million people with life and disability
income insurance, retirement services, and investment products such as
mutual funds, securities, variable life insurance, and variable annuities. The
company also supplies employee benefits programs to six million
participants, including life, health, and dental insurance, as well as qualified
pension plans. In addition to regional home offices in New York City;
Bethlehem, Pennsylvania; Spokane, Washington; and Appleton, Wisconsin,
the company has 55 remote sales offices and 80 remote agency offices.
Like other insurance companies, Guardian Life is an information
intensive organization where data processing and communications network
infrastructure have consistently been important contributors to its success.
Guardian Life’s IT organization has earned numerous accolades including
multiple CIO100 awards from CIO magazine [PRNE11]. According to Dennis
Callahan, Executive Vice President and Chief Information Officer for,
Guardian Life, “A strong partnership between IT and the businesses enables
Guardian to deliver cost-effective technology services that facilitate world-
class customer service, product innovation, and operational efficiency.”
Ensuring alignment between business and IT is important to Guardian Life
and provides a consistent theme for many of the insurance companies IT
projects including its data center consolidation initiatives [CIOZ12].
Data center consolidation has been an ongoing concern at Guardian for
more than a decade. Guardian’s IT governance structure is team-oriented
and the company’s data center consolidation initiatives are overseen by it
Infrastructure team. The Infrastructure team is primarily co-located in New
York, and Bethlehem, Pennsylvania but it has key support teams in Spokane,
Washington, Appleton, Wisconsin, and Pittsfield, Massachusetts.
Guardian Life began taking a serious look at data center consolidation in
2000, but in the aftermath of the September 11, 2001 terrorist attack,
Guardian also became more concerned with business continuity issues.
Guardian had four significant data centers, at its four home offices, but the
primary data center was in New York City. After 9/11, Guardian wanted
make infrastructure changes to ensure business continuity across its existing
data centers and made plans to add two more data centers to the mix.
Guardian performed an assessment of its data centers to provide a basis
for planning on the location of data processing resources. One surprising
outcome of this assessment had to do with utilization. The assessment
revealed that the four data centers had about 1000 UNIX and NT servers,
with an average capacity utilization of 10%. Even at peak demand, only 25%
of the processing power of the servers was being used [MUSI02]. Guardian
responded to this assessment with a plan that included the following
1. Move the primary data center from New York City to Bethlehem, Pennsylvania.
2. Improve the efficiency of its data centers, including server and storage utilization. Specifically, Guardian set a goal of reducing the number of servers supporting Guardian’s applications and databases by 40% and reducing the server support staff by 60%.
3. Ensure a smooth transition to the new primary data center.
The company chose IT consulting firm Greenwich Technology Partners
(GTP) to help it design and carry out the transition. GTP began with an
assessment of the company’s IT environment and looked at the impact of
moving the data center from New York to Bethlehem. A major issue related
to the move was that the largest number of data center users was located in
the New York area. Thus, the new deployment needed to provide sufficient
data transmission capacity to meet these users’ needs. Fortunately, the
network infrastructure already in place was fairly standard and easily
scalable (Figure C7.1).
An ATM WAN backbone linked the four regional home offices. Frame
relay connections linked the remote sales offices and remote agency offices,
and 100-Mbps and 1-Gbps Ethernet LANs provided connectivity within
campuses. The Cisco Catalyst 5500 and 7200 series routers provided
Ethernet support combined with a modular, easily scalable design. The Cisco
IGX 8400 ATM switches could be scaled to support an ATM network service
at any desired capacity. With three widely used networking technologies in
place, some of the problems that might have been encountered in a more
convoluted networking environment were avoided.
GTP also looked at the application and database patterns. They
determined that in addition to traditional applications such as file and print
services, PeopleSoft, and Lotus Notes, Guardian also used a collection of
applications to support its intranet. Then, as now, the site included
marketing materials and sales tools for the firm’s agents, account profiling,
and customer data. The company had also invested in a number of financial
services applications, including applications for supporting its trading and
securities functions. Due to their complexity and the amount of resource
required to support them, many of Guardian’s applications have historically
been supported by dedicated servers.
The transition team, consisting of Guardian and GTP personnel, did
extensive validation work and benchmarking to make sure the data they had
gathered during the initial assessment were accurate. They measured
network utilization at granular levels and modeled various consolidation
scenarios for reducing server hardware.
From this analysis, GTP proceeded to develop a plan for consolidating
the servers, looking at both business and technology issues. For example,
the team considered the criticality of the applications supported by the
servers, as well as which business units they belonged to. Some servers
were good candidates for consolidation; others were not; and others were
out of warranty, which made them too expensive to keep. After the
migration plan was devised and the new architecture developed, more
testing was conducted to ensure their viability.
The team devoted much thought and analysis to the migration to the
new architecture, so as not to disrupt day-to-day operations. The plan
targeted the least complex opportunities first, starting with file and print
services. Guardian initially had more than 30 servers providing file and print
services. These were consolidated into just two servers clustered in a high-
availability, fault-tolerant configuration. For more complex parts of the plan,
the team opted to do some of the consolidation in New York, and only then
move the servers to Bethlehem after the consolidation had settled down.
The initial consolidation and relocation project yielded tangible benefits
to Guardian in terms of reduced hardware and personnel requirements. But
the benefits extended well beyond these initial objectives. The mindset of
solving new problems efficiently and in the context of the existing
infrastructure had taken hold. Guardian no longer automatically takes orders
for new servers to support applications as it did throughout the 1990s.
Instead, Guardian analyzes each new application requirement and attempts
to support it with the existing hardware/software suite or with minimal
upgrades and extensions.
The total cost of the data centered consolidation project was $4.5
million, but the company saved more than $3 million in 2002, offsetting
much of that cost. Even greater savings were realized during each of the
following two years.
Further Consolidation In 2010, Guardian embarked on a second major data centered consolidation
initiative. The maturation of virtualization technologies and evolution of high-
bandwidth WAN connections encouraged Guardian to consolidate six data
centers into two [MITC11]. This is illustrated in Figure C7.2. Guardian will
consolidate its mission critical data processing infrastructure into one
primary data center that it will own. The company plans to lease a second
modular pod for use as its second data center. Pod data centers can be
thought of as data centers in a box. These can be configured by vendors to
customer specifications and delivered as a container that looks similar to
refrigerated box car or multi-modal shipping container stacked on ships or
carried by tractor trailer trucks. Pods are energy efficient but can be a
cramped for human movement.
In addition to moving to a much leaner data center infrastructure,
Guardian is migrating from Unix to Linux. It has also started to move some
of its applications to the cloud [MITC11]. By the beginning of 2012, Guardian
had moved 18 back-end applications to SaaS and had begun transitioning e-
mail, HRIS, and IT services into the cloud. The breadth of Guardian’s move
to the cloud put the company on the leading edge among Fortune 250
Guardian is using the IDEAS Advantage services from Ideas
International to help it make decisions about remodeling its data center
infrastructure [GOLI11]. These sophisticated tools in combination with
Guardian’s commitment to cloud services indicate that the insurance
company’s data center consolidation initiatives are far from over. Guardian
has not yet moved its core ERP systems to the cloud, but even this is under
Discussion Points 1. Do some Internet research on the reasons why businesses generally
invest in data center consolidation projects. What benefits do they commonly hope to realize? How do Guardian’s rationale for consolidating data centers compare to those of other businesses?
2. Getting outside consultants to manage data center consolidation
projects is a common practice. Discuss the pros and cons of using consultants to manage data center consolidation projects?
3. How/why has virtualization fueled business interest in data center
4. Why is the availability of high-speed, high bandwidth communications an important consideration in data center consolidation plans and decision-making?
5. Do some Internet research on modular (pod) data centers. Summarize
the advantages of disadvantages of modular data centers.