AWS re:Invent 2025 - It's About Time! Improving Distributed Systems with Amazon Time Sync (CMP409)

Improving Distributed Systems with Amazon Time Sync

Understanding Clocks and Time Synchronization

• Clocks are fundamental components in computing systems, consisting of a periodic signal and a counter
• Hardware clocks can vary in accuracy, size, and cost, ranging from small crystal oscillators to large atomic clocks
• All hardware clocks drift over time due to factors like temperature, power supply, and aging
• Network Time Protocol (NTP) is commonly used to correct clock drift by synchronizing client clocks to a reference server clock
• However, network latency and asymmetry introduce challenges in achieving accurate time synchronization across distributed systems

AWS Approach to Precision Time Synchronization

• AWS developed the Amazon Time Sync service to provide highly accurate time synchronization for EC2 instances
• This service uses dedicated hardware clocks and GPS signals to deliver sub-microsecond accuracy at the software layer
• The Nitro system in EC2 instances provides direct access to these hardware clocks and time sources, bypassing the network
• This allows for hardware packet timestamping with nanosecond precision, enabling improved observability and ordering of distributed events

Leveraging Precision Time in Distributed Systems

• Accurate clocks enable simpler distributed system designs by reducing the need for complex coordination mechanisms like distributed locks and leader election
• Applications can leverage the precise time information to order events and messages with high confidence, improving consistency and reducing retries
• Examples:
  • Yugabyte DB saw a 1000x reduction in retries by using the Amazon Time Sync service and Clockbound
  • AWS databases like Aurora DSQL and DynamoDB Global Tables use the precision time capabilities to ensure consistency across regions

Clockbound: A Simplified Time Synchronization Solution

• AWS has released the Clockbound open-source software, which provides a standalone time client and daemon
• Clockbound abstracts the complexity of time synchronization, providing applications with the current time, the uncertainty window, and the clock status
• It uses a "feed-forward" design optimized for cloud environments, rather than traditional feedback-based time synchronization
• Clockbound achieves sub-microsecond accuracy, with a typical uncertainty window of 20-26 microseconds

Business Impact and Real-World Applications

• NASDAQ is leveraging the precision time capabilities to enable the deployment of their high-frequency trading systems on public cloud infrastructure
• By using hardware timestamping and the Amazon Time Sync service, NASDAQ can maintain the deterministic and fair order processing required for their exchanges, even in a distributed cloud environment
• Other AWS partners, like Yugabyte DB, have seen significant performance improvements and reduced complexity by integrating the Amazon Time Sync service and Clockbound into their distributed systems

In summary, the AWS approach to precision time synchronization, including the Amazon Time Sync service and Clockbound, provides a powerful set of tools for building highly accurate and performant distributed systems. By offloading the complexity of time management, AWS enables customers to focus on their core business logic and leverage the benefits of cloud computing, even for the most demanding low-latency and high-throughput applications.