The following article was contributed to OUT-LAW by Owen
Cole, Technical Director UK & Ireland, F5.
There’s been a lot of hype over the past year surrounding
'green' computing and the drive to lower the impact of IT and data
centres on the environment. While we’re all for the concept of
green computing and reducing the impact of computing on our
environment, we’re also cognisant of the reality that every IT
organisation also has to worry about the other kind of
green: its bottom line.
The good news is that there is some amount of overlap between
these green computing initiatives. Reducing power consumption and
management expenses, and increasing efficiency of existing
resources through consolidation and virtualisation decreases both
the impact of devices on the environment as well as on IT’s
increasingly tightening budget.
Reducing power and heat
The easiest way to reduce the impact of any device on the bottom
line, be it a server or networking equipment, is to reduce the
amount of power it requires. Modern servers often draw variable
amounts of power based on the processing power in use by
applications.
Similarly, some networking equipment and other devices provide
the same functionality, drawing varying amounts of power based on
their load and configuration. This can be beneficial in reducing
the operating cost of the server or device, but like dealing with
variable costs of bandwidth due to bursts in usage, also makes it
difficult to estimate annual costs and budget appropriately.
Another simple but often overlooked facet is how many BTUs
[British Thermal Units] are generated by any given device. By
decreasing the BTUs generated, there is less heat and thus less
cooling required within the data centre.
The costs of cooling a data centre are larger than those to heat
one, owing to the fact that much of the heating needs in a data
centre are inherently taken care of by the BTUs generated by the
devices it houses. Reducing these costs can have a significant
impact on the operating expenses of any IT organisation.
Reducing power consumption and generation of BTUs for devices
and servers is something over which IT has no control. While IT can
certainly use such ratings as part of its decision making process
for purchasing, it really can’t do a thing to affect how much power
is consumed or how many BTUs are generated by any given device.
It’s simply a cost of doing business.
Yet IT can make decisions, both in purchasing and
architecture, which reduce power consumption and heat generation by
reducing the number of servers and devices that make up its data
centre. Consolidation and virtualisation are both ways in which IT
can positively impact its bottom line.
The impact of consolidation
Consolidation has been an ‘initiative’ in IT for many years, and
it generally revolves around the consolidation of the data centre
in terms of the number of servers deployed to support
mission-critical applications. While reducing the number of servers
in the data centre, and thus rack density, both power consumption
and heat generation can be positively affected.
Yet capacity needs must be balanced with consolidation efforts,
and at some point consolidation is no longer possible. As the
volume of users and application usage grows, so must the number of
servers – and devices such as application delivery controllers –
necessary to scale mission-critical applications.
Striking a balance between scalability and controlling costs is
difficult, and thus far it has been nearly impossible to avoid the
deployment of additional application delivery controllers as a
mechanism for scaling out a data centre.
Whether chassis or appliance-based, these devices have only
added to the cost of power consumption and increased the generation
of heat within the data centre, raising operational costs.
Solving this problem requires effort on both the part of the
application delivery controller vendor to reduce the power
consumption and BTU generation of devices while simultaneously
providing a way to scale without increasing the number of devices
required for deployment within the data centre.
A single, chassis-based application delivery controller
requiring less power and generating fewer BTUs that also scales via
a virtualised bladed architecture can address the growing need for
capacity without adversely impacting IT’s bottom line, or the
environment.
The impact of virtualisation
By architecting a new breed of chassis-based application
delivery controllers that take advantage of virtualisation not only
at the server level but at the chassis and blade level, these new
devices can provide better performance in a single unit than could
previously be obtained with multiple appliance-based solutions or
legacy chassis-models.
By virtualising blades and CPUs, essentially creating a single,
powerful processing matrix, this new breed of chassis-based
application delivery controller can scale nearly linearly.
This internal processing scalability means that every last drop
of processing power is being used and can provide a much higher
capacity than its legacy ancestors. By more efficiently using the
processing power available, the performance per power unit is
increased, making each transaction processed by the application
delivery controller cost a fraction in terms of power consumption
than would otherwise be possible.
Figure 1: Comparison of performance per watt for
legacy and new chassis model application delivery
controllers
| |
Layer 7 CPS |
Watts |
CPS per watt |
BTUs |
| Legacy chassis |
58,000 |
4,620 |
12 |
15,763 |
| New chassis model |
1,260,000 |
1,463 |
862 |
4,991 |
Consider the comparison in Figure 1. Regardless of what the cost
per kilowatt hour, there is a significant saving in terms of power
when moving from the legacy chassis-model to a new, virtualised
chassis-model. This has a significant positive impact on the
environment as well as on the organisational budget.
Given the higher performance capacity of the new chassis model,
this also allows for fewer devices necessary to meet the growing
traffic management and application delivery needs of today’s IT
organisations, which lowers the cost of operations as well as
management.
The management costs of such a new breed of application delivery
controller are inherently lower than a traditional application
delivery solution, owing to its virtualised architecture and the
ability for the device – and IT manager – to manage the system as a
single entity rather than as individual blades in a larger
system.
This reduces the amount of management necessary, and in turn
reduces the costs associated with managing the device.
This is especially true as capacity is added, as it would
require multiple legacy chassis-based devices to match the
processing power of a single virtualised chassis-based system. Each
added device must be managed, and adds to the amount of power
consumed and BTUs generated, making it much more expensive to
scale.
Also having an impact are the BTUs generated by each device.
There is a definitive cost associated with removing the heat
generated by these devices in the form of cooling, so the lower BTU
generation of the new breed of chassis-based solution is a definite
boon both on the environment as well as on the budget.
Conclusion
It’s rare that an environmentally-friendly movement such as
Green IT results in reducing costs, especially in its early stages.
And yet in the case of this new breed of chassis-based application
delivery controllers, that’s exactly the result. With the decreased
management and power consumption costs and increased performance,
these new application delivery controllers are both green as in
grass and green as in cash.
F5 is exhibiting at Storage Expo, 14th
– 15th October 2009, Olympia, London.
