Can weather forecasts protect your data?

When Perth hit high temperatures above 44°C in January 2015, one of Australia’s largest internet service providers, iiNet, decided to shut down some of its systems at its data centre as a “precautionary measure”. This effectively led to thousands of disgruntled customers who were cut off from the internet for over six hours.

The company wrote in a Twitter post: “Due to record breaking temperatures, iiNet Toolbox, Email and our corporate websites are unavailable. Apologies for any inconvenience caused.”

One angry customer reportedly tweeted back: “@iiNet We can’t get internet. Glad you’re communicating reasons, but not sure “it’s hot” is really an acceptable excuse in Australia.”^[1]

According to the State of the Climate 2014 report released by the Bureau of Meteorology (BOM) and CSIRO, our climate has warmed by 0.9°C since 1910, with more frequent extreme heat events and fewer cool extremes.

Australian temperatures are projected to continue to increase, while the duration, frequency and intensity of heatwaves have increased across large parts of the country (and we can probably safely assume will continue).

In light of this, the angry customer’s comment therefore opens up an interesting question in relation to data centre performance in the face of changing weather patterns in Australia. Should we be better prepared, and is there a way for businesses to ensure they avoid equipment damage, downtime and customer dissatisfaction, in the face of these more extreme weather events?

Keeping your cool

Schneider Electric Australia Operations Manager Weather Services Matt Saunderson believes there is — by using the rather precise weather forecasting models now at our disposal.

“This is a relatively new concept [for data services]. In the past, data centres have tended to rely on monitoring conditions themselves or receiving notifications of government issued warnings,” he said.

In the US, Schneider Electric’s weather monitoring service is relied on by energy utilities, agriculture, transportation agencies as well as for preparation around big outdoor sporting events. The company kicked off the service in Australia in 2015 and is planning towards a similar uptake in the region for businesses seeking to avoid weather-related issues.

Schneider Electric uses the DiCAST forecasting system, which integrates a range of different computer models and data sources to create the most accurate forecast based on the recent performance of each of those components.

“We can then apply input from meteorologists over the top of that process to ensure optimum accuracy and performance. This technology was developed by NCAR (National Centre for Atmospheric Research) in the US,” explained Saunderson, adding that forecasting computer models are ever-evolving and getting more precise.

“The proven accuracy of our forecasts, combined with evolving capabilities of data centres to take proactive measures when weather threatens, sees this just being available now.”

Effectively managing high and prolonged heat of course is important for the lifespan of data centre equipment, since a temperature elevation today could cause damage and therefore unexpected system failure in the future. When you get to a third or fourth day above 40°C, that shouldn’t be the time to think about what to do when the UPS fails and the cooling system is in overdrive.

Being alerted ahead of time will help with planning your airflow cooling requirements.

Weather alerts

If you’re thinking about taking advantage of weather forecasting to manage external temperature variances, Saunderson said data centre managers can start with installing a weather station to take readings.

“Schneider Electric’s weather offer provides complete information for business customers that have some exposure to the weather,” he said. “The primary offer is our SaaS offer, called WeatherSentry Online, which provides an online weather decision support system to weather-sensitive businesses.”

This system integrates all key weather data sources, and can include hyper-local forecasts based on a customer’s weather station. A Google Maps-based interface, which allows integration of all key datasets that can be tailored to specific user groups, includes weather radar, precision lightning strikes, weather observations, satellite data, localised forecasts and a range of other data types and weather hazard information.

Email alerts, push notifications to smartphones and SMS messages are available for significant conditions observed, on top of daily forecasts up to 15 days out or hourly forecasts up to 72 hours. A 24x7 online consulting service with meteorologists based in Australia, who can also provide additional tailored content, is also available to help managers make better decisions for their data centres.

More than a reaction

Weather forecasting offers businesses an alternative to just reacting to extreme weather events as they happen. However, what if loads could be seamlessly shifted between different data centres to where cooling is more efficient and costs might be lower in the event of high temperatures? Saunderson said this concept is currently being looked at being developed between the data centre experts in Schneider Electric Australia and the weather team.

“This would see two potential benefits: increased reliability, as load is shifted away from servers that are in high temperature areas and at increased risk of power supply problems or cooling system problems; and reduced cooling costs, as load is shifted to areas where free-air cooling could be utilised, rather than incurring the expense of artificial cooling systems,” explained Saunderson.

Image courtesy Schneider Electric.