How the Boston Celtics used cloud to up their basketball analytics ...

When Boston Celtics CTO Jay Wessland began working with the National Basketball Association team in 1990, the age of big data was still a ways off — especially in the professional sports arena.

It would be six years before digital data was cheaper than paper and another six until Major League Baseball’s Oakland Athletics rode analytics through the storybook season writer Michael Lewis documented in his 2003 book “Moneyball: The Art of Winning an Unfair Game.”

Modernization for the Celtics in the 1990s meant collecting basic data on a courtside computer, a practice that took hold only after Wessland arrived. When the now 78-year-old NBA franchise appointed Wessland director of technology in 2000, the IT hadn’t advanced much further. The organization promoted him to CTO in 2016.

“I couldn't imagine why a pro sports team would even need a CTO but, in today's world, everything revolves around the tech stack and technology,” Wessland said last month, during a virtual event marking the near-completion of his organization’s six-year journey to cloud.

Professional sports, like any other big business, is a data-driven, IT-dependent enterprise, where elastic compute and scalable storage deliver a competitive edge — on and off the court. Cloud modernization has become table stakes, from banking to basketball.

“It's now way past an SQL server and a web server that I could build myself,” Wessler said. “We need a lot of help to do this.”

The Celtics moved quickly to cloud in 2017, choosing AWS as its primary provider. The initial lift-and-shift took months, not years, according to Jonathan LaCour, CTO at Mission Cloud, the AWS consulting service that managed the migration.

“Technically speaking, 99% of the infrastructure was moved out of on-prem,” LaCour said in an interview with CIO Dive.

That’s when the real work of refactoring legacy applications, rationalizing data estates and optimizing the tech stack began. “Modernization and refinement have been happening continuously ever since,” LaCour said.

The data tipping point

Digital transformation isn’t a one-and-done endeavor. Untended technical debt haunts cloud deployments, undermining efficiency and ballooning costs. Organizations that lack a sound cloud business strategy or migration plan can’t squint hard enough to see the ROI.

“People can get into a situation where they do a migration and then just let it sit there and they become unsatisfied” LaCour said. “If you’re running your data center workload in a cloud, that’s not the way to save money.”

Wessland had a relatively small IT group and decades of data to migrate from infrastructure he’d built from the ground up.

“We ran everything on Windows in our own data center — if you could call it a data center,” he said. “It was actually the room behind my office with a couple racks.”

Data was the tipping point.

As sports and business became more analytics driven, Wessland’s SQL servers strained to keep pace with the workloads. Lack of computing resources on the back end translated to a competitive disadvantage.

Data from the West Coast wouldn’t arrive in time to be processed before the next game, Wessland said. Coaches would arrive at the team’s facility after a night game and ask Wessland’s team for reports that they lacked the technology to produce fast enough.

“The prior infrastructure was what you would call startup infrastructure,” said LaCour. “They bought a lot of off-the-shelf software to fulfill their use cases and focused on hiring statisticians, data analysts and data scientists.”

Cloud talent was not something Wessland had on-prem.

“They knew they couldn’t be the AWS experts but they wanted to get to the cloud,” LaCour said.

A 4-phase migration

By 2017, Wessland was running an analytics shop. The SQL servers he’d built were straining to keep pace with the workloads and with the pace of the NBA.

Migrations have to start somewhere but they also need a long-term roadmap. Mission Cloud settled on a four-part plan for the organization, outlined in a case study published last month.

Wessland described the initial lift-and-shift to AWS as a proof-of-concept to demonstrate the organization could take what was on-prem and scale it in public cloud. The migration addressed one of the Celtic’s biggest problems — speed and scalability — almost immediately.

“The Celtics have a very bursty workload,” LaCour said. “They also had a very specific challenge to solve: We have big iron in Jay’s office, which does allow us to vertically scale, but we run into limits and it takes us 14 to 15 hours to process the data. It's an absolutely perfect use case for the cloud.”

The heavy lifting began in the second phase, when Wessland retired his MS SQL servers, moved from Windows to a Linux-based operating system and spun up serverless containers.

“The most challenging thing was taking that first leap from MS SQL and Windows,” said Wessland. “Just thinking about it made our heads explode. It was hard but we got through it.”

The team eventually settled on Snowflake for data and single cloud for everything else.

“Our experience is that the vast majority of our customers have no interest in being multicloud because it significantly complicates their use case,” LaCour said.

Post-migration maturity

As the organization’s cloud strategy matured, distinct migration segments blended into incremental process and technology enhancements. Phase three, as LaCour describes it, involved “leaning into what's available in AWS,” including automation, resiliency, developer efficiency and security features.

“We’ve separated the modernization into big chunks, and sometimes it happens like that,” Wessland said. “But more often than not, we're making little changes all the time.”

The most recent pivot — phase four — was to multiple AWS accounts, giving different parts of the business a place to land. Wessland’s biggest workflows are basketball analytics, he said, but other deployments have grown up around the business, including enterprise analytics and IT infrastructure.

Dividing accounts by function provides visibility into who’s spending how much on what kinds of workloads, Wessland said. “It's also helped our data security know that the business analytics guys that are worried about ticket revenue aren't in the same place logging into the same systems as the basketball analytics guys,” he added.

The days of trying to scrape together enough compute to churn through game-planning data are in the past.

“Maybe we get into the playoffs and some coach wants us to crank out a crazy plan he's got for how to guard LeBron [James] and it might take a bunch of data,” Wessland said. “We can spin the machines, do it and then spin them back down and it might cost us $100 overnight. We don't have to run them forever — just until we get the answer we’re looking for.”