- Andrew
Des’Ascoyne, Director at Spook Limited (
Powering
Up IT Availability or Playing Jenga in the Data Centre
Energy consumption has become a priority focus for
any IT Manager. From escalating costs to
carbon footprint concerns, gaining insight into power utilisation is key. But
organisations are overlooking the fast rising business risk associated with
power overloads. With overstretched IT teams delivering more innovation and
change than ever, data centre design creep is endemic. The result? The vast
majority of disaster recovery invocations in the past year were power related.
Companies cannot afford to play Jenga with their IT
resources; blithely hoping that each new server can be added without, finally,
tripping the circuit breaker and taking out the whole data centre. It is time
to leverage the insight provided by power monitoring tools and understand just
how power is being used across the infrastructure.
As Andrew des’Ascoyne, Director at Spook explains,
with proactive, predictive analysis based on real time power monitoring,
organisations can radically reduce the risk of power related incidents and
undertake highly effective power based load balancing to maximise performance
and minimise energy consumption across the data centre.
Design
Creep
Every IT Manager knows the importance of effectively
managing data centre power consumption.
From the drive to reduce Power Usage Efficiency (PUE) ratings to
corporate CSR pledges, the need to reduce energy related expenditure and
address Energy Efficiency Scheme CRC requirements, data centre power
consumption is now centre stage. Organisations are increasingly exploring
innovative real-time power monitoring and metering tools to assess energy
consumption and highlight areas for improvement – from introducing new cooling
technologies to creating more efficient data centre design.
But while power consumption is a serious issue – and
cost – many organisations are simply too tightly stretched to impose control
over power usage in the data centre to minimise overload and ensure the
underlying power resources are used effectively and safely. As a result, the most beautifully designed
data centre can descend into chaos within a matter of months.
New servers are added onto racks, overloading one
phase coming into the server room and creating an imbalance between phases; or
old servers are swapped for newer models without considering the best location
on a power, as well as space, utilisation perspective. Yes, the new server might be physically
smaller than the old one, which means, in theory, it could be repositioned to
maximise space – but is the underlying power infrastructure adequate? And what
about the air conditioning: although a new server may demand less power for
equivalent processing capability, it will typically generate more heat than the
old kit, creating a situation where the existing air conditioning is not
sufficient for the new heat footprint.
Power
Led Invocation
The result of this lack of understanding or
forethought is an increasingly devastating effect on business. According to recently published data from
SunGard Availability Services, some 27% of invocations in 2010 were power
related. Indeed the trend in power
related incidents has increased steadily – while it has been listed as a
contributor for invocation in the last 15 consecutive years, it appeared as the
main reason for invocation in 2010.
Add to that the consideration that 80% of data
centres do not have adequate or, in some cases, any back-up power generation
facility and the potential financial cost of such disruption is significant. A
CA Technologies’ study reveals €17 billion in revenue is being lost in each
year in the time taken to recover from unplanned IT downtime. It also comments
that “post-downtime” there is an additional delay during which time data is
still being recovered and that during the post-outage period, company revenue
generation is still severely hampered, down by an average of 25%.
Predictive
Model
So what are the options? Measuring and understanding
power consumption is key to reducing power related disruption; minimising the
risk of overload through changes to planned maintenance, for example, and using
real time alerts to ensure problems are rapidly and proactively addressed
before downtime occurs. By understanding the way power is used within the data
centre 24x7, organisations can ensure proactive load balancing activity that
ensures consumption is well distributed - preventing the problems caused by
overload can then become a standard procedure.
Of course, given the nature of IT resources,
undertaking this process in-house is probably not realistic. Gaining insight
into power consumption and the implications of design change needs to be made
easy: organisations simply do not have the time, expertise or, to be frank,
desire to manually consolidate meter readings from multiple Power Distribution
Units (PDU) to gain a complete overview. By linking real time monitoring with a
single view of the entire data centre’s power utilisation, backed up with
simple weekly and monthly reports that highlight both risks and opportunities
for improvement, organisations have immediate insight into power utilisation
and load balance. With this insight an organisation can begin to impose greater
control and ensure staff do not simply use the nearest plug or PDU but
understand the dangers of power overload and the value of load balancing,
further reducing the risk of data centre design creep.
Conclusion
Balancing thin resources and the business demand for
constant change, the vast majority of IT Managers know that it is just a matter
of time before a new development – either long term planned or to resolve an
immediate problem – will be a step too far; the piece of Jenga that trips the circuit and takes down an entire
data centre. Is it really worth either the risk or the ever present stress
when, by taking a proactive approach, it is possible to both understand power
usage and, critically, predict the impact of data centre change
Those organisations that are simply monitoring the
data centre to check energy consumption figures are missing a trick and
seriously overlooking the risks associated with power failure. Metering is
becoming a fundamental component of data centre design and performance but
organisations need to proactively use this information to minimise IT risk as
well as improve efficiency.
great information
ReplyDelete