Five ways to kill data centre efficiency

27 November 2015 | Mike Bennett

Cover

Mike Bennett

Blog Author | CenturyLink; VP of global data centre acquisition and expansion, EMEA


As the telecoms industry becomes more digitised – from the emergence of IoT technologies to fully digital ecosystems – telecoms providers will rely more heavily on data centre reliability to support their businesses. However, the benefits of digitisation, such as customer satisfaction, enhanced revenue streams and cost savings, are no longer applicable if your data centre is poorly managed.

As the telecoms industry becomes more digitised – from the emergence of IoT technologies to fully digital ecosystems – telecoms providers will rely more heavily on data centre reliability to support their businesses. However, the benefits of digitisation, such as customer satisfaction, enhanced revenue streams and cost savings, are no longer applicable if your data centre is poorly managed.

Approaching this from a reverse brainstorming perspective, by answering “How can I cause a problem?” reveals findings that can actually be very helpful in avoiding or solving the problem.  With this in mind, we’ve decided to look at the top ways to kill data centre efficiency and provide some top tips; just remember to reverse the tips, if it’s efficiency you’re after!   

 

1. Just get the thing built

If you’ve decided to build and kit out your data centre all in one go; good you’ve already bypassed modular build efficiencies in deployment, energy efficiency and high density. When it comes to location, stick it anywhere. Slough is known as the “data centre capital” of the UK due to its superb connectivity and easy access along the M4 corridor, so is one place to be certain to avoid. 

But, if you’re lumbered with a well-designed and nicely built facility in a good location, do not despair, the vast majority of inefficiencies can be achieved in management and operations (M&O). 

 

2. Forget risk assessment

You’re in the United Kingdom; if you’re not on a floodplain or earthquake fault line then surely the biggest “risks” you face are heat waves of 30 degrees – what other risks can there possibly be? Sure, the leading cause of reported data centre outages are directly attributed to shortfalls in management, staff activity and operation procedures (as recognised by the Uptime Institute) but when searching for inefficiency, thorough risk assessment and carefully considered mitigation measures should be tossed aside.

 

3. Wild west it

How the IT load is set up and maintained on the data centre floor, by those co-locating, has a big impact on energy efficiency and other users. Therefore, make sure to discourage good data centre citizenship. The simplest things should be ignored here; from the positioning of equipment in racks to maintaining hot and cold aisle integrity, and you can completely forget best practice procedures when cabling up servers so as not to impede airflow, and the use of blanking plates. Avoiding these details will help drive inefficiency on the ground. Just consider the impact of not just one neglected server but thousands – that’s a lot of airflow to be throttled!

 

4. Embrace passive racks

Passive racks that are not consuming any power are great for taking up valuable data centre space without achieving anything. Stanford University and consulting firm Anthesis Group suggests there are 10 million physical servers currently inside data centres around the world that are not actually being used. You can help increase that number by not building to the required density. 

Not putting the load in the asset is also a bonus; if a service only needs and uses 50kw, why not allocate 500kw (just in case!)? It’s all extra inefficiency!

 

5. Don’t be too picky

The people inside the data centres are the real differentiators, they can make a good data centre great but they can also make a potentially great data centre horribly inefficient. 

Data centres are complex beasts at a mechanical and electrical level and there are a myriad of things that can cause an outage or problem, from patching racks, through to testing generators and failing to test the air conditioning. Failure to understand a service’s configuration or what’s going on in a suite is a key part of ensuring chaos and requires a lack of appropriate training. 

Regardless of whether you’re running the entire data centre yourself, or just dropping in to handle the upgrades and changes, you’ll want to avoid training and professional development. The Uptime Institute M&O Stamp of Approval should be avoided at all cost, for it provides an outcome-based guideline that centres on operational excellence. The independent researcher and certification body has analysed 20 years’ worth of data centre outages and therefore should be ignored, if you’re hoping to repeat inefficiencies.

 

Conclusion

Of course, all these tips are in jest, but the process of reverse brainstorming helps illustrate what is truly important when driving efficiency.  By examining how easily inefficiency can be brought about, you’ll ensure happier customers and improved margins overall.