Executive Summary

• Idle servers are burning up to 30% of your power budget and it’s time to hunt down those idle loads to free up capacity.
• Fortune estimates that 10 million servers are idle, meaning $30 billion in wasted capital.
• PCE is crucial to give us the awareness of how much compute is used (and wasted) to enhance efficiency, promote sustainability, reduce environmental impact and comply with regulations, whilst maintaining cost-effective operations.

You can have a fancy plan for your capacity and seek out the latest cooling upgrade to ‘cope’ with the demand, but have you considered idle loads eating up your capacity and wasting power?

On average, your servers are burning up to 30% of your power budget doing diddly squat. They are essentially ghosts, costing you in more ways than one and on the road to sustainability, it’s time to exorcise them.

You don’t have to crack out your grey zip-up suit and your Proton pack to bust these ghosts, but you do have to address the problem hiding inside your servers if you want to free up stranded capacity (and let’s face it, why wouldn’t you).

An idle server still burns 40-70% of its peak power

Even at single-digit utilisation, servers draw roughly half of their maximum load and whilst there are recent designs that reduce this, this is still a major problem in the face of today’s power constraints and efficiency battle. If modern servers drop to a  ~10% utilisation, that doesn’t mean a 90% reduction in power consumption still remains around 50%. This is for one idle server, so across a typical enterprise averaging 20-30% utilisation, the cost of idle servers compounds into a substantial amount, which is a hidden waste stream that can be fixed. Fortune estimates that 10 million servers are idle, meaning $30 billion in wasted capital.

Exorcising your ghost servers

Say “Efficiency” three times in the mirror and your problems are solved. Kidding. To exorcise your high-maintenance ghost servers, it’s all about enhancing efficiency and making your infrastructure more sustainable. First of all, this starts with being painfully aware of just how efficient (and inefficient) your infrastructure actually is right now. PUE is a good starting point, but you can’t rely on that alone to help you determine your efficiency because it doesn’t address how effectively power is turned into usable compute. You could have a fabulous PUE score, but you may not be able to deploy additional compute because you’re still power-stranded and limited. Even with the same PUE score, two facilities may have different levels of usable capacity because it depends on how much power is converted into IT load.

Enter stage right: PCE, also known as Power Compute Effectiveness. It measures how effective a data centre converts power capacity into active IT load, showing you exactly how much power is available and what’s being used (and what’s idle).

Using both metrics provides a complete view of how your data centre is performing and gives you an insight into just how many of your servers are ghosts.

It’s an allocation and orchestration problem

Once the metrics have made the road clearer as to how much compute is being wasted, it’s time to look at your allocations software to manage compute more effectively. It’s not a problem you build your way out; it is about orchestrating your power capacity to ensure it’s being used efficiently, and that’s all about the software you use (and the smarter software you need) to maximise resources.

If we’re power-constrained now, imagine the landscape if we didn’t deploy resources to exercise our idle ghost servers. The adoption of PCE is crucial to enhance efficiency, promote sustainability, reduce environmental impact and comply with regulations, whilst maintaining cost-effective operations.