Met Office Makes a Cost-Conscious and Test-Driven Decision to Consolidate Linux Instances on the Mainframe
Richard Cains (left), technical lead with Met Office’s mainframe team stands with Martyn Catlow, portfolio lead for centralized IT infrastructure, outside the Met Office headquarters in Exeter, Devon, England. Photo by Simon Keitch
|Customer: Met Office
Headquarters: Exeter, Devon, England
Business: Weather modeling and forecasting, and climate-change research
Challenge: Choosing between a mainframe refresh or maintaining a more distributed Linux computing environment
Solution: After taking costing and benchmarking test results into consideration, the company moved many of its formerly distributed Linux workloads to an IBM zEnterprise 196
Hardware: Two IBM zEnterprise 196s with 17 IFLs each
Software: Oracle database software
You check the weather forecast on a Monday and see it’s going to be sunny all week. Alas, on Wednesday, it’s raining. That’s the curse of weather forecasting. Occasionally, no matter how many models are run and how good the science, probability runs foul of the physics.
But that’s quickly changing, thanks to the ever-increasing power of supercomputers and much more sophisticated modeling capabilities. One organization that’s taking up the challenge of improving forecasts for both its commercial and noncommercial customers is the Met Office, a provider of global weather forecasts, based in the U.K.
“By consolidating distributed commodity servers, however, you can save a great deal of money. And believe you me, when we looked at all of these parameters, it just made sense to move the workload to the mainframe.”
—Martyn Catlow, Met Office portfolio lead for centralized IT infrastructure
Not only does it have a supercomputing environment to power its weather modeling, but also two IBM zEnterprise* 196s (z196s) that help propagate that forecasting information to end users and customers. Much of this is done thanks to the organization’s heavy and increasing use of Linux* running on a series of Integrated Facilities for Linux (IFLs). As is often the case with this type of consolidation, Met Office is now in the position to better serve its downstream customers and contain operating costs—not to mention more accurately tell people whether they need their umbrellas or not.
Based in Exeter, Devon, England, Met Office has been operating since 1854, first providing forecasts for ocean-going ships. When aviation became prevalent, it expanded its portfolio to serve aircraft as well. After World War II, Met Office began using early computers to improve its forecasting. In the late 1960s, it introduced its first IBM mainframe to its computing lineup.
Now it offers forecasting information and weather-related products to a variety of customers. “We forecast for the public and a wide range of commercial sectors, and have a strong history of forecasting for the marine and aviation sectors,” explains Martyn Catlow, Met Office portfolio lead for centralized IT infrastructure. “We also produce weather products for defense and a wide range of retail and infrastructure customers, such as national road and utility services.”
Notably, global climate change research is a large aspect of Met Office’s predictions and a huge consumer of IT capabilities, Catlow says. “The data produced by this activity to date amounts to around 17 petabytes and is archived in a specially designed, multi-storage-tier data warehouse,” he adds.
The heavy-duty number crunching involved in the creation of numerical forecasts is conducted in a supercomputing environment based on IBM POWER7* systems. It analyzes a complex range of parameters, including the obvious temperature, pressure and humidity. The forecast models are run four times a day, with the most critical being at midnight and noon when the most observational data is available.
comments powered by