The concept of workload portability is often raised by vendors in very conceptual ways – the idea of being able to move workloads is great, but reality of both practical and use-case issues is a little more difficult. Given this situation I was interested to hear about a case study being circulated by cloud portability vendor CliQr, detailing how one company is making use of portability in very real ways.

Pratt and Miller is a full service engineering company, and is apparently a well-recognized name in racing circles with victories in American Le Mans series, SCCA Pro Racing World Challenge, NASCAR, IndyCar and the GRAND-AM series. The company provides design, development, construction and at-the-track engineering support to racecar makers like Corvette, Cadillac, and Chevrolet.

Like many organizations with the double issues of limited compute resource and spiky demand – Pratt & Miller was faced with the challenge of insufficient compute resources that were reducing their ability to run simulations, costing them valuable time and potentially a race. One of Pratt & Miller’s key software products is a Windows-based vehicle simulation called Pratt Miller Lap Time Simulation (PM-LTS), which gives racing teams a vehicle model to predict lap times. The software accepts input parameters for all aspects of the vehicle and race track being simulated including vehicle information such as tiers, brakes, suspension, engine and drive train; and race track details including elevation, banking, friction, cornering and atmospheric conditions. Using these variables, Pratt and Miller ran simulations on a single machine, with each simulation often exceeding an hour to complete, and if multiple iterations were required, the run time increased to days.

Anyway – given the various parameter of what they do, the simulations became a bottleneck and the company decided to run simulations across multiple infrastructures. They used CliQr to create a template for their application, thereby allowing for migration across different infrastructures – Pratt & Miller was able to transfer simulation files from local machines to the cloud and run the simulations on cloud compute – all the while retaining the same basic application stack and form. Abstracting the stack away from the infrastructure it sits upon, the company is also able to run unmodified on any public cloud provider – therefore being able to make decisions based on factors such as cost without having issues around compatibility.

In terms of bottom line results, Pratt & Miller was able to increase processing power anywhere from two to hundreds of times with no changes to the base application. In the past Pratt & Miller users ran simulations on a single multi-core machine and a base simulation model took one hour to run and 200 iterations typically took multiple days. Moving to the cloud, users are now able to run iterations in parallel, enabling 200 iterations to run in about 1.5 hours.

As I said, this idea of workload migration is often spoken of in entirely conceptual terms, or worse conflated to live cloudbursting. While a use case such as simply moving workloads between on-prem and public cloud infrastructures isn’t a particularly sex story, it’s one which delivers real value. It’s also really refreshing to see how these cloud migration products allow smaller organizations, that wouldn’t be able to justify re-writing applications to run on the cloud, to leverage the benefits and scalability the cloud can bring.

Ben Kepes

Ben Kepes is a technology evangelist, an investor, a commentator and a business adviser. Ben covers the convergence of technology, mobile, ubiquity and agility, all enabled by the Cloud. His areas of interest extend to enterprise software, software integration, financial/accounting software, platforms and infrastructure as well as articulating technology simply for everyday users.

Leave a Reply