Has Putin started a server revolution?
The European Commission wants Member States to reduce consumption. “Demand reduction is fundamental: it lowers energy bills, ends Putin’s ability to weaponise his energy resources, reduces emissions and helps rebalance the energy market,” executive vice-president, Frans Timmermans said in a statement accompanying the Commission’s proposals.
Following on from the German government’s plans to reduce its own energy usage, Deutsche Bank unveiled its own measures in August. Among them are plans to reduce energy usage in “technical rooms.”
Undoubtedly, the UK government will be looking closely at how the rest of Europe copes. But irrespective of whatever discussions come out of the Cabinet and Number 10 in the coming weeks, heads of finance in every business will be looking at ways to cut their fuel bills dramatically.
And, as can be seen from what Deutsche Bank has already said, server rooms and datacentres will be among the big energy guzzlers that will need throttling back.
Since the dawn of blade servers and virtual machines, the IT sector has recognised that it’s not easy getting the megawatts of electricity needed for new datacentre builds. Thankfully, servers have become more efficient and utilisation – the proportion of actual usage compared to idle time – has increased. Gartner analyst Tiny Haynes puts server utilisation at around 70%, but there’s room to increase to upwards of 90%. Investing in systems management and ensuring server resources are not ring-fenced around a particular project’s budget, makes it easier for server admins to increase utilisation. In doing so, they can move workloads to servers with more spare computing capacity, reducing the need to purchase and power up new server resources.
Energy requirements to achieve an outcome
Taking a system-wide approach is something else we should all consider. Even if a modern processor consumes more watts of electricity than older processors, as McKinsey partner, Ondrej Burkacky, points out, it can actually do far more work. For instance, graphics processing units (GPUs) are notoriously power hungry. One only has to look at the tech spec-sheet for high-end gaming machines to appreciate just how much electrical power GPUs need. But, when running workloads designed for a highly parallel architecture, the GPU gets things done an awful lot quicker. The individual components may be power hungry, but the overall energy consumption required by the workload can be significantly less than if it was run on a traditional PC server.
Arm is now making a big push in the server space. As it cements its credentials, it too, will offer a compelling alternative in those businesses doing the sums to balance energy usage with workload system requirements. What we are now seeing is the next evolution in datacentre computing, which is being accelerated by soaring energy bills brought on by the war in Ukraine.