Today's business environment runs 24 hours a day.
Customer demand and the "follow the sun" necessities of global
operation mean that access to transactions and services is needed
24 hours a day, seven days a week, and downtime has massive
business impact.
Meanwhile,
data-retention legislation means that businesses are forced to
store ever-larger volumes of information.
Green concerns mean they have to do this while minimising
energy consumption. As if that was not enough, businesses also
expect to be able to react quickly to their markets and rapidly
roll out software to support new business processes.
Because datacentres are under intense pressure to perform with
exceptional reliability, good design and management are vital.
Where once the datacentre was a bunch of servers in a room,
pressures on space, power supply and cooling capacity - plus the
need to reconfigure quickly to meet new business demands - mean
that design has to be extremely well thought out.
At the same time, hardware and software assets must be managed,
and the requirement of business continuity must be dealt with. So,
what are the key planks of datacentre design and operations that
can ensure 24x7 operations?
First, it is worth stepping back a little and defining what the
datacentre is meant to do so that we can determine its key
characteristics.
The key aim of any datacentre should be to cost-effectively
support the technology needs of the business, says James Staten,
principal analyst at Forrester Research.
"What this means to datacentre design is making sure the right
level of facility is matched to the needs of the business," Staten
says.
"For example, if the business requires the datacentre to provide
24x7, high-performance internet-based application availability,
that has strong implications on the reliability, availability and
serviceability features of the datacentre, and how many datacentres
are needed to accomplish this goal. If the demands are less, the
integrity of the datacentre can be lower."
The key purpose of a datacenter is to support the operations of
the business. So it is critical for the IT department to work side
by side with business to understand strategy so that the
infrastructure can scale with the business.
At the same time, the datacentre has to be able to change to
meet the demands of the business as the market environment changes.
Simplicity that can enable flexibility and agility needs to be
built in from the start, says Guy Bunker, chief scientist at
Symantec.
"The design needs to be able to react quickly and efficiently to
changing business needs. Time equals money, and by creating a
standardised infrastructure, management costs can be reduced.
Complexity is the enemy of the IT administrator, and no more so
than in the datacentre," Bunker says.
For that reason, datacentre experts recommend building as much
standardisation and modularity into the hardware inventory as
possible. This makes sense because when compared with high-cost
silos, standardisation allows a greater degree of commonality and
more rapid change to meet business needs.
Modularity of hardware can also help to optimise power and
cooling, which is a vital consideration for the datacentre, as not
only do efficient servers cost less, they also perform more
reliably, says Ian Brooks, vertical marketing manager for the UK
and Ireland at Hewlett-Packard.
"Indications show that it already costs more to power and cool a
server over its lifetime than it does to buy the server. With
multiplying numbers of servers, higher densities and hotter
processors, it is clear that IT facilities are running out of
cooling capacity and power. This is a problem regardless of
platform - rack, tower, blade - all datacentres need to address
it," Brooks says.
Power and cooling problems in datacentres are a product of the
increasing power densities of modern IT hardware. Introducing the
most efficient power distribution and cooling mechanisms will
result in the lowest cost to the organisation for running the
datacentre.
Traditional raised-floor datacentres using forced cool air can
no longer provide enough cooling power for the maximum power
densities, says mark Blackburn, chief technologist with management
software supplier 1E.
"A 19-inch rack full of blade servers would require, on average,
eight perforated floor tiles in front of it to maintain the
requisite airflow to keep them cool," Blackburn says.
"Advances in CPU efficiencies with the advent of multi-core and
smaller die sizes through further miniaturisation of logic
circuitry alleviates this somewhat, but the trend for higher and
higher densities forced by Moore's Law leads to requiring new power
and cooling designs to increase efficiency."
Innovations in cooling
There are a number of innovations in cooling that provide
alternatives to computer-room air chillers, such as using ducted
cool air from the outside environment, or using water-based heat
exchangers.
Making your datacentre a lights-out environment can also cut
cooling requirements. Using remote access, you can ensure doors are
not opening and closing all the time, which means the air
conditioning does not have to work as hard to keep the room
temperature within acceptable limits.
Cooling is also aided by rearranging your racks to create hot
and cool aisles, says Bill Allen, director of product management
with remote access software supplier Avocent.
"Position your racks so that one row has the front of equipment
facing the other and the next row is the backs of the racks facing
inward," he says. "This can help with airflow, especially if the
air conditioning ductwork overhead is aimed directly on the 'hot'
aisles. This exercise will also force you to be smarter about
cabling issues which can impede good airflow," he says.
It is also possible to reduce power consumption and cooling
needs by making sure applications are running with the correct
number of servers. Blackburn says, "Many applications are
over-served, and turning off unused capacity can be a quick and
easy win. Aslo, turn off unused systems – there are many legacy
systems that are no longer utilised that can be decommissioned,
thereby saving power."
Metered and switched power-management devices also help match
power to actual processing needs. That way you can power down
specific, non-critical, machines during "off hours" to lessen the
power draw on each rack.
Powering up machines in a sequence instead of all at once can
also help prevent a massive power draw at the boot-up moment.
After dealing with the physical design and layout of the
datacentre, the next step is to consider managing the datacentre
and its hardware and software assets.
At the highest level, where the computing infrastructure meets
the physical infrastructure, Blackburn recommends an interface
between the IT and facilities departments.
"Organisations should handle the datacentre holistically, and
integrate the IT and facilities staff by bringing them under a
single management structure," Blackburn says.
"Making the person responsible for buying the servers also
responsible for paying the electricity bill will ensure that more
thought is given to datacentre efficiency as a whole. Attempt,
wherever possible, to integrate building-management and
systems-management tools. This gives you an insight into how
operations are impacting power consumption and it is then possible
to make small changes to systems operations."
Then there are the software tools that can help manage the
datacentre's hardware and software. It is possible to get tools
from suppliers such as HP, Tivoli and BMC that provide
single-screen, real-time supplier agnostic views of:
● Monitor power and cooling requirements
● Failover and redundancy for disaster prevention
● System health monitoring and alerting for disaster
prevention
● Audit logs and reporting for company- or government-mandated
compliance requirements
● Access and control of virtual servers as well as physical
servers.
Such management tools may sometimes provide aids to change
management too, although point products are also available. Key
tasks these tools assist with are provisioning of new server
environments, for example.
Automating server provisioning can have a dramatic impact on the
length of time it takes to get a new system operational, and can
provide many benefits in large, complex, interdependent
environments, especially in the test and development environments
where many servers may need to be repeatedly rebuilt.
Keeping track of changes
Another vital aid to change management is the configuration
management database (CMDB), says Symantec's Guy Bunker. "A CMDB is
rapidly becoming an essential element in tracking and controlling
systems in the datacentre.
"This can be coupled into an automated provisioning system, so
that machines, especially blades, can be quickly and efficiently
re-provisioned and therefore re-used on multiple tasks. CMDBs are
also essential in tracking 'configuration drift', which becomes
critical when looking at disaster recovery planning," he says.
Closely allied with change management is preventative
maintenance, which nowadays is more about software than hardware.
Hardware has achieved mean time between failure figures in the
hundreds of thousands or even millions of hours, so good
preventative maintenance is nowadays synonymous with good patch
management.
This can be handled fairly easily within the datacentre through
planned downtime and a good change and configuration management
system.
Last but by no means least there is the training and skills
aspect of datacentre management. Here, IT departments need to
ensure they have effective Information Technology Infrastructure
Library- and IT Service Management-compliant procedures in place.
These provide industry best practices and standard operating
procedures in terms of IT deployment that can be adapted to any IT
organisation.
Simplifying your environment and judicious use of outsourcing
can also help, says Quocirca's Clive Longbottom.
"Keep it simple. Consolidate to a single transport such as TCP/IP,
rationalise applications and services, provide first-level support
yourself – such as forgotten passwords – and outsource more
technical aspects," Longbottom says.
"Also think about moving basic technology skills into more of a
business-technical translator capability to bridge the gap between
the business and the IT capability."