Hawk-Eye Innovations Powers Real-Time Sports Data with Flink and Amazon MSK

This blog is co-authored by Brent Evans, Gergely Jahn, Alex Manning, and Marco Lorenzi, from Sony Sports Businesses.

The world of professional sports has entered a new era where continuous data streams power every aspect of the game from officiating to fan engagement. The real-time tracking technology of Hawk-Eye Innovations, a Sony-owned company and AWS Strategic Partner, has become the backbone of modern sports analytics. Their sophisticated camera networks process up to 480 messages every second, delivering millimeter precision insights within milliseconds. This streaming infrastructure enables near real-time (NRT) officiating decisions, powers immersive broadcast experiences with augmented reality overlays, and provides teams with continuous biomechanical data for performance optimization. However, this infrastructure was straining to keep up with the growth of Hawk-Eye and needed to change.

Existing infrastructure and operational challenges

Current Infrastructure Overview

As Hawk-Eye continues to expand their sports technology offerings globally, their infrastructure was built around a self-managed Apache Pulsar deployment running on Amazon Elastic Compute Cloud (Amazon EC2) instances. This architecture was designed to handle their core NRT sports analytics workloads with variable processing demands of 15-125 MBps for each game depending on the complexity of sport. The infrastructure supported diverse sports including football, basketball, baseball, soccer, tennis, and hockey.

The existing system was also architected to meet strict sub-second latency requirements, while supporting multiple concurrent events across different time zones.

Infrastructure challenges and pain points

As the global footprint of Hawk-Eye expanded, this existing infrastructure began facing increasing operational strain across several key areas:

• Operational burden of self-managed services: Managing their own Apache Pulsar infrastructure internally created significant operational overhead and became a major pain point for the team. This diverted valuable resources from core innovation activities.
• Unnecessary costs from static resources: the self-managed Apache Pulsar deployment of Hawk-Eye on EC2 instances required maintaining fixed compute capacity to handle peak loads, resulting in over-provisioned resources during periods of low activity. The static infrastructure lacked the ability to dynamically scale processing resources up or down based on real-time demand. This became particularly challenging when handling sudden spikes in concurrent matches across different sports events and time zones, impacting both operational efficiency and cost optimization.
• Complex stream processing development: The existing architecture required significant development effort for new features and sports offerings, making time-to-market less predictable and increasing development costs.
• Integration complexity: As part of the broader Sony Sports ecosystem, Hawk-Eye needed to enable easier integration capabilities for other divisions, while maintaining their high standards for real-time performance.

These challenges led Hawk-Eye to seek a more scalable, managed solution that could support their continued innovation while reducing operational overhead and improving cost efficiency.

Solution

Hawk-Eye has implemented a robust real-time data processing architecture leveraging the managed streaming services of Amazon Web Services (AWS) to address these challenges.

Main AWS services the solution uses are:

• Amazon Managed Streaming for Apache Kafka (Amazon MSK): A fully managed service that makes it quick to ingest and process streaming data in real-time with Apache Kafka. Amazon MSK handles the heavy lifting of managing Kafka infrastructure while providing the reliability and throughput needed for mission-critical sports applications.
• Amazon Managed Service for Apache Flink: Designed to perform computations at in-memory speed and at scale, this fully managed, serverless experience helps organizations to run Flink applications without managing infrastructure or clusters. It’s ideal for real-time stream processing where low latency is crucial.
• Amazon Simple Storage Service (Amazon S3) with Apache Iceberg: Provides scalable object storage with advanced table formats that enable schema evolution and efficient querying of large datasets. It is perfect for storing and analyzing historical sports data alongside real-time streams.

Figure 1: AWS architecture for real-time sports tracking data stream.

Solution overview

Following is a walkthrough of the four main sections of the architecture.

1. Data ingestion layer:
   1. On-site infrastructure with Hawk-Eye cameras streams video to the cloud-hosted tracking services.
   2. The tracking services analyze the video to determine player and ball positions, then convert this data into protocol buffer (protobuf) messages—a compact binary format designed for efficient data transmission.
   3. These protobuf messages are then ingested into Amazon MSK for further processing and delivery to downstream systems.
2. Stream processing layer:
   1. Job1 processes streaming data for real-time analytics and outputs to Amazon OpenSearch Service for live dashboard visualization.
   2. Job2 enriches and transforms raw tracking data into standardized formats and feeds processed data back to Amazon MSK for downstream consumers.
   3. Job3 prepares data for long-term storage, aggregates tracking information, and stores processed data to Amazon S3 for on-demand analytics.
   4. This layer scales Job1, Job2 and Job3 to meet consumer-specific tracking data requirements.
3. Storage layer:
   1. A dedicated Flink job processes frame-level data and buffers write to Amazon S3 to optimize storage costs.
   2. Processed data is stored in Amazon S3 using Apache Iceberg table format for efficient storage of historical tracking data, support for on-demand analytics, and schema evolution capabilities for changing data requirements.
   3. OpenSearch Service provides real-time indexing and search capabilities for live feed requirements, while Iceberg on Amazon S3 offers cost-effective long-term storage with ACID (Atomicity, Consistency, Isolation, and Durability) transactions and time-travel queries for historical analysis.
4. Tracking data consumers:
   1. Multiple downstream applications use the tracking data for various purposes, including creating exciting experiences for sports fans, providing live game statistics, monitoring player health and safety, and measuring player performance.
   2. Various applications consume the processed tracking data to support quick access for on-screen graphics and overlays during live broadcasts.
   3. The data also feeds internal machine learning model training for improved player analytics and predictive insights.

Benefits

Since instituting the solution architecture they have experienced multiple benefits.

The architecture verifies:

• Enterprise-scale messaging with 40-50% better performance.
• Multi-consumer support enabled 80% reduction in consumer lag through optimized partitions.
• Scalable storage provided 50% storage cost reduction with Amazon S3 integration.
• Flexible integration with over 200 connectors enabled 60% faster deployment.
• Operational efficiency with 80% less management time and 60% total cost of ownership reduction.

Conclusion

The migration to AWS managed services eliminated operational overhead while maintaining up to 480 messages each second processing requirements for sports analytics. The multi-consumer architecture enabled separate streams for real-time processing, graphics systems, and external data distribution without performance degradation. It also provided serverless capabilities with automatic scaling that optimized costs across varying workloads (15-125 MB/sec for each game). Integration of Apache Iceberg with Amazon S3 established a robust foundation for historical analytics while maintaining sub-second latency for live operations.

The solution’s modular design created a scalable foundation that streamlines integration for other Sony Sports divisions and enables rapid deployment of new sports offerings. This architectural approach confirms the system can adapt to future requirements, while maintaining high-performance standards necessary for professional sports broadcasting. It positions Sony Sports, and Hawk-Eye, for continued growth in the competitive sports media landscape.

Contact an AWS Representative to know how we can help accelerate your business.

Further reading

• Amazon MSK resources
• Best practices for Standard and Express brokers
• Understand MSK Connect
• Maintain best practices for Managed Service for Apache Flink applications
• Security in Amazon Managed Service for Apache Flink

About Hawk-Eye Innovations

Hawk-Eye Innovations, a Sony-owned company and AWS Strategic Partner, has revolutionized sports technology with its precision tracking and decision-making systems. At its core, Hawk-Eye deploys sophisticated networks of specialized cameras that capture sports action with remarkable accuracy. As a best-in-class sports technology provider, Hawk-Eye’s vision is: to pioneer change in sport, challenging convention for its betterment through a unique blend of creativity, tech, and innovation. With over 1,000 team members operating from offices across 43 countries, Hawk-Eye delivers groundbreaking technology that enhances fan experiences worldwide.

From SkeleTRACK to player and ball tracking, Hawk-Eye’s performance tracking data provides a fully-rounded story of sports action. This technology has been a cornerstone of sports officiating innovation, used in everything from tennis Grand Slams to tier one soccer and American leagues. In addition, it generates world-class datasets, which are stored, aggregated, delivered live and on-demand to sports technology partners for analytics and immersive experience.