5 ways big data won the America’s Cup for Ainslie and Oracle Team USA
6 min read
27 September 2013
Oracle Team USA's Asim Khan tells us how 300 sensors and 3,000 variables helped one of the greatest sporting comebacks of all time.
Technology played a bigger part in this year’s Americas Cup than ever before. Thousands of spectators both shore-side and at home were able to watch live on-board HD footage of the competitors as they battled in the bay, and the on-board technology was second to none.
Building and sailing an AC72 catamaran demands extensive performance analysis based on collecting huge amounts of data, and applying the right analytics to improve boat design and performance.
Here are the five essential tools that helped Oracle Team USA to succeed:
1. Data collection
- More than 300 sensors throughout the boat collected a huge amount of performance data and transmit it to a server in the hull;
- About 3,000 variables were running at ten times a second – all while sailing – from sensors that measured strain on the mast to angle sensors on the wing sail that monitored the effectiveness of each adjustment;
- Several video feeds and numerous still images of the sail wing were captured every second; and
- About a gigabyte of raw data was pulled per boat per day as well as about 200 gigabytes of video per day.
Depending on what we’re measuring, we’ll configure a light feed of about 150 key parameters and transmit them in real-time to the Oracle Database on the performance chase boat. For example, if we were measuring sail wing performance in certain conditions, the feed will be heavy on data from the wing.
2. Real-time analytics
- The performance chase boat is the analytical hub, where we ran a whole raft of different analyses live while the boat is sailing. Sometimes the analysis requires a very complicated combination of ten, 20, or 30 variables fitted through a time-based algorithm to give us predictions on what will happen in the next few seconds, or minutes, or even hours in terms of weather analysis;
- We relied on the Oracle Database, along with a number of screens around the boat running simple Java applications on displays that showed us the sensor-generated numbers;
- We ran a four-man crew, all involved in real-time analysis geared to make the AC72 really scream; and
- One team member analysed data from the sail and wing, while another looked for data trends; a system technician monitored the system itself, and I looked at the data from a sailor’s point of view to see how the performance was changing. We can also pull data from the shore system via a 4G connection.
3. Performance sailing technology
- Each crew member wore a ruggedised PDA on his wrist and received a real-time, customized feed of information to help improve performance—what the load balance is on a particular rope, for example, or the current aerodynamic performance of the wing sail; and
- We didn’t want to overburden the wireless network, but at the same time foiling demands requires real-time, accurate information so we used Java for the PDA displays—it’s lightweight and could maintain a high refresh rate for up to 30 devices without crashing the network.
4. Historical analysis
- We used a custom application called Race Cutter to package historical data into a geographical frame for review;
- The design and sailing teams were able to compare today’s sailing data to information from a specific point in time and analyse any number of performance factors—the strain on the dagger boards, or the load on the rope, for example;
- We also had still photos and video frames linked to that data, so when you clicked on a certain point in time, it jumped to the images and video from a number of different camera viewpoints; and
- It was all linked to the live database information coming into Race Cutter, which was backboned by the Oracle Database.
5. Extreme database performance
- We upgraded from our previous hardware to an Oracle Exadata Database Machine – the performance improvements were startling; and
- Exadata gave us roughly a 10 times speed improvement on CPU-intensive tasks, meaning that when dissecting a training run for example, we were able to get at the data while the sail was still fresh in the teams head, so every second of performance improvement translates to better support for the sailing team.
The result could not have been closer: a 9 – 8 win over an incredibly talented Emirates Team New Zealand. Big data may not have been a crew-member, but there’s no doubt of the vital role it played in that performance.
Asim Khan is director of information systems at Oracle Team USA.