Tags: available space, business continuity, center managers, center space, constraints, core research, density servers, disaster, dublin, floor space, gartner, hosting companies, new hardware, perspectives, quality data, rakesh kumar, rapid uptake, ras, rationalization, space power,
A Message From Data Center Managers to
CIOs: Floor Space, Power and Cooling Will
Limit Our Growth
Gartner RAS Core Research Note G00142393, Rakesh Kumar, 21 August 2006 R1617 03302007
ost data centers will struggle to
M accommodate the growth, as well as the
power and cooling requirements, of new high-
Most large businesses are reaching the limit of their
available space and are struggling to host new
equipment. Turning to hosting companies is also
density servers, which will result in an inability to proving to be difficult because they are experiencing
meet growing business needs. CIOs need to be a rapid uptake of their real estate. In Europe, for
aware of data centers' potential limitations. example, London, Dublin and Amsterdam have
limited quality data center space available for renting.
ANALYSIS As a result, most users are trying to find creative
During the next three years, most CIOs will ways of hosting the new boxes and sometimes settle
experience constraints in data center floor space and for office locations that do not meet data center
power that could limit an IT organization's ability to design criteria and are not best suited from disaster
grow as the business grows. The problem is the recovery and business continuity perspectives.
result of two dynamics:
· As organizations buy more hardware, they need The problem, however, is not only about available
more floor space to house it. Although companies data center space. Most of the new hardware is
are going through server rationalization programs, designed to an increasing level of density, which may
net data center floor space is increasing at 5% to appear to reduce unit floor space (for the individual
10% a year. server) but needs a considerable amount of power
and cooling. Moreover, the space around these high-
· The emerging server and storage technologies are
density boxes has to be greater than with the
highly dense and require an exponential increase
previous generation of servers because of the need
in power and cooling. Running these systems in
to channel cool air around the aisles.
legacy data centers may not be possible for many
organizations because these old facilities have not
For example, rack power greater than 20 kilowatts
been designed for the emerging power
(kW) approaching 30 kW in the next 18 months
requirements.
per rack can be needed for a high-density blade
(HDB) environment compared with a rack power
For many companies, building new facilities will be
consumption of less than 2 kW per rack for a non-
expensive and is not core to their business.
HDB environment. A similar amount of electricity is
Therefore, CIOs must realize that the infrastructure
needed to remove the huge amount of heat
used to accommodate IT growth may become the
generated from these machines. As the number of
bottleneck.
racks goes up, the problem increases exponentially.
There may be a need to bolster air conditioning units
Ability to Grow Your Organization's IT (which is not always possible) and strengthen the
Capabilities floor to carry the heavier load. Thus, many facilities
Servers worldwide grew at approximately 12% more cannot host the number of new servers that are
in units in 2005 than 2004, and this figure will needed for the business growth.
probably increase during the next few years. Raw
storage growth will maintain a base level of between CIOs may opt to build new data centers designed to
30% and 70% compound annual growth rate. As new host more equipment, have a greater amount of free
service-orientated application development takes off, space and be more resilient; however, this is
these numbers will increase significantly.
problematic as well. The Uptime Institute defines four example, the IT hardware market is dynamic, with
levels of data center reliability. Tier 1 is the most changes in a number of areas. Demand for hardware
basic level and Tier 4 is the most resilient. The cost naturally increases. More importantly, the refresh
of a Tier 4 data center can be two to three times the rates for incumbent components and appliances are
cost of a Tier 2 center. Data center build costs have increasing rapidly. New technologies are being
increased by more than 20% during the past three introduced at a faster pace than ever before (such as
years. More importantly, newly designed data servers, storage, PDAs and processor virtualization).
centers will need to have power requirements of With typically an excess of players in the server,
approximately 5 million kW to 10 million kW. Not only storage and network component areas, hardware
would the build costs be beyond most IT budgets, prices are falling, so it often makes sense for users
but the ongoing operational costs would create a to change equipment quickly. This creates a problem
huge imbalance on annual IT budgets. Energy costs for introducing new technologies into data centers.
account for less than 10% of most IT budgets. Although the speed of new hardware procurement
However, this could rise to more than 30% as new can be measured in weeks, any changes to a data
systems are used. Google, for example is among a center facility (a new data center or refurbishing an
small but significant number of organizations whose existing site) are measured in years. If this demand
annual energy bill is bigger than their annual server for new technology outstrips a data center's
budgets. Gartner believes that more than 50% of infrastructure to house and support that technology
large organizations will experience a similar change quickly, then not only will this affect the perception of
during the next three years. IT, but it could create a competitive disadvantage for
the organization. Thus, new data centers need to be
Securing such a supply of power would not be easy designed in modular ways so they can scale in line
for many large cities. Local governments may with new hardware growth.
impose limits or tariffs, or there may not be enough
supply. One way around this is the development of Space vs. Power It All Adds Up
new data center behaviors. For example, we are As the density of systems (for example, servers,
seeing the early stages of new models, such as joint storage and networking hardware) increases during
ventures between a large user's IT organization and the next five or more years, the footprint of each unit
a hosting or outsourcing player. Such an approach may well decrease, but the energy requirements will
would allow the huge upfront and operational costs go up. This will lead to a trade-off between data
to be shared by both organizations and, thereby, center space to host the equipment vs. the cost of
secure the development of such a new facility. the energy to power and cool the technology. A rack
However, there are dangers. Working out occupancy of HDBs may seem to provide more computing
and expansion plans for a five-year period in the power per square foot of floor space than a rack
initial contract requires complex future growth optimized system, but it will require significantly more
planning in a relatively uncertain world. Also, only a power and cooling. It will probably also require more
limited number of players in the market would feel floor space to house the server because of a greater
comfortable with such a solution because of need to distribute the heat generated throughout a
governance and legal reasons. wider area. Also, the incremental power and cooling
demand will increase the support space for computer
For CIOs to plan for IT growth, agility and the speed room air conditioning units, higher-capacity
of introducing new technology are important. Unless uninterrupted power supply systems and generators.
new data centers are designed in a modular way Thus, the space savings will be largely illusory.
that enables speedy expansion, they will not able to
change fast enough to meet user requirements. For
© 2006 Gartner, Inc. and/or its Affiliates. All Rights Reserved. Reproduction and distribution of this publication in any form without prior
written permission is forbidden. The information contained herein has been obtained from sources believed to be reliable. Gartner
disclaims all warranties as to the accuracy, completeness or adequacy of such information. Although Gartner's research may discuss
legal issues related to the information technology business, Gartner does not provide legal advice or services and its research should
not be construed or used as such. Gartner shall have no liability for errors, omissions or inadequacies in the information contained
herein or for interpretations thereof. The opinions expressed herein are subject to change without notice.
2
The economics of data center real estate will play a Recommendations
significant role. In most U.S. markets, industrial CIOs need to quantify and manage the problem of
zoned space is inexpensive: $12 to $15 per square new infrastructure growth. Many methods can be
foot net vs. energy costs of $60 to $80 per square used in large data centers, and all of these should be
foot. Unless the client is located in expensive real employed in a coordinated manner:
estate, density is problematic from a cost standpoint.
· Improve the use of existing equipment because
Spreading things out mitigates the problem. However,
this will delay the need to move to high-density
in industries such as banking, large data center
systems.
facilities exist in the heart of expensive locations,
such as Manhattan or London. CIOs will, therefore, · Balance the technical requirements of new
be faced with a difficult equation based on many equipment with power and cooling specifications.
different parameters. For example, a data center in a · Configure data centers to balance air flow
large city like London will have high real-estate costs between hot and cool zones, and create special
(potentially limiting the introduction of new high-density zones for more-demanding
technology) and will require adequate supplies of infrastructure. Install chassis that use chilled water
power. City planners are more likely to impose and explore other solutions as they become
energy surcharges and regulate energy waste in such available during the next couple of years.
locations. A new data center in a remote location will
be more affordable because of land prices, which
could allow a wider spread of servers in the facility.
However, such a site may incur high networking costs
and it may be difficult to attract skilled labor to such
locations. In some cases, the data centers may be
best suited close to power stations.
3