Measuring usefulness

The data centre represents an enormous investment for any business – not only during the initial design and build phase, but also to ensure that operations work optimally on an ongoing basis. Effectively utilising the sheer number of IT assets the data centre holds has consequently become one of an organisation’s biggest challenges. What should be a logical and straightforward assessment of how IT assets, such as racks and servers, are being used in the data centre has, over time, manifested itself in overcomplicated metrics.

While great strides have been made in improving server and power management levels, it is the industry’s ineffective metrics that fail to provide the full picture when it comes to understanding exactly how efficient a data centre really is. By looking beyond traditional measurements and without starting a complicated and expensive Data Centre Infrastructure Management (DCIM) project, and instead focusing on utilising those that are more business-relevant as well as seeking to examine data utilisation more closely, organisations will find scope for greater efficiency.

HOLISTIC ENERGY EFFICIENCY
When contemplating the efficiency of the data centre, organisations tend to concentrate most of their efforts on the amount of power being consumed, predominantly by the servers that form the basis of the data centre IT infrastructure and subsequently how best they can minimise the amount and cost of energy. Traditional metrics, such as PUE (Power Usage Effectiveness) and DCIE (Data Centre Infrastructure Efficiency), are used universally to discover how much energy is consumed compared to the energy that a data centre actually requires. However, simply monitoring and managing energy levels in this way can give a false indication of how a data centre is operating, because it only a provides a surface view.

In the contemporary data centre, the vast amounts of energy consumed account for a significant part of the IT budget, while the resulting carbon emissions have become especially important given the stringent government regulations concerning carbon reduction. With this in mind, it is easy to see why organisations predominantly focus on these kinds of energy measurements when trying to improve on data centre efficiency.

The problem is that energy metrics like PUE and DCIE are limited in their scope because they look wholly at usage while ignoring the efficiency or business value of what is happening on the servers. Fundamentally, solely focusing on energy saving metrics ignores the holistic view of the data centre.

TOWARDS BUSINESS RELEVANCE IN DATA CENTRE METRICS
The Corporate Average Datacentre Efficiency (CADE) metric developed by The Uptime Institute and the 451 Group seeks to determine the efficiency of both data centre facilities and IT assets as well as of efficiency modulated by utilisation. Many have criticised CADE for being too difficult to fully comprehend, difficult to compute and producing vague – and even counter intuitive – results. However, CADE attempts to provide a measurement that can work as a metric for ‘useful work’, by calculating a percentage of the usefulness of the server.

An alternative is the recently released Fixed to Variable Energy Ration (FVER), which has been conceived by the British Computing Society (BCS). FVER suggests that a ’proxy’ is selected for the data processing activity and used to represent the value of this investment to the business. A proxy can be factors such as file transfers, the number of downloads or the number of transactions.

Whilst these two metrics have attempted to highlight business-related performance, the results they have produced have proved too abstract to be properly employed and, crucially, still do not give a clear insight into how the data within the server is being utilised.

THE BLACK BOX
At the heart of the issue is what business value servers perform – it is not about the level of business value, high or low, but about whether your servers are doing anything useful at all. If there is no useful activity, then your investment is giving you absolutely no return.

Therefore, when it comes to measuring true efficiency, intelligence is needed about actual server usage which demonstrates the types of processes taking place on a server as well as taking into account that these operations are required by, and provide value to the business. Research shows that servers use about 60 percent of their maximum power while doing nothing at all and typically only run at around 15 percent utilisation.

With IT assets and server maintenance taking up a significant chunk of an organisation’s budget, it is imperative to have a metric that goes beyond measuring the average CPU utilisation. What CPU won’t show is whether activity is due to value-producing business computing or background administrative, errant or stuck processes. Whether this activity is useful to the business or not is determined by guesswork and inference. Essentially, data centre professionals are limited to observing the ’vital signs’ of the server in question, not the processes occurring within it.

While the data centre community has long acknowledged that this type of analysis lacks critical details, it simply knows of no other alternative. A lack of real intelligence into what servers in the data centre are actually doing has also, in part, led to organisations being hesitant in decommissioning servers that they suspect are carrying out little or no activity, for fear this may disrupt critical systems or interrupt the business process. Such uncertainty has only served to create ineffectual efficiency processes.

REAL INTELLIGENCE
The solution for data centre operators lies in making the shift from vague approximations, assumptive percentages and abstract proxies, towards deploying next generation server assessment tools that give data centre operators real clarity and visibility into data processing activity. Only with these types of tools can organisations gather the intelligence required in order to categorise server activity as useful or non-useful; critical or non-critical.
Those servers and/ or processes which are identified as surplus to requirements can then either be safely decommissioned or repurposed without fear that this will disrupt the business process or affect critical systems. The knock on effect of reducing server number creates more costs savings in terms of space, power, cooling, software licensing, software maintenance, hardware maintenance and operations support.

These types of metrics can be invaluable for data centre consolidation projects and virtualisation initiatives, enabling IT managers to carry out aggressive efficiency strategies with no risk and at a very low cost. Clear visibility into the data processing activity across the server estate can enable high levels of cost savings, capacity planning, and asset value realisation.

Whilst categorisation of server activity is not the only aspect that needs to be considered when driving for greater efficiency, it is nevertheless a largely neglected aspect of the server management and efficiency process and is also very quick and easy to do with no disruption to everyday business processes.