AWS re:Invent 2025 - Maximizing uptime: Itaú’s mission-critical mainframe migration to AWS (IND3304)

Itaú's Mission-Critical Mainframe Migration to AWS

Introduction

• Itaú Bank is one of the largest banks in Latin America, with over 70 million customers and a 100-year history.
• Itaú recognized the need to modernize its core banking systems to keep up with changing customer behaviors and digital transformation in the financial industry.
• The bank decided to migrate its mission-critical checking account platform, which had been running on a mainframe for over 50 years, to the cloud on AWS.

Key Challenges and Requirements

• Ensure high availability, with a target of 99.99% uptime (less than 5 minutes of downtime per month).
• Maintain absolute accuracy and consistency of customer account balances, a non-negotiable requirement.
• Handle massive scale, with the ability to authorize up to 6,000 transactions per second at peak times and support large accounts with 1,000 transactions per second.
• Minimize the impact of failures, ensuring that only a portion of customers are affected rather than all customers.

Architecture: Cell-Based Design for High Availability

• Adopted a cell-based architecture using the Kámadávara (KVA) framework, which provides isolation, fault tolerance, and scalability.
• Each cell is a fully autonomous unit that owns its own data and can operate independently.
• The router layer directs customer transactions to the appropriate cell based on a partitioning strategy (e.g., full table mapping, prefix range, consistent hashing).
• Each cell has multiple replicas (active-active or active-standby) to ensure high availability, with replicas placed in different availability zones.
• The journal component coordinates data replication across cell replicas using a quorum-based approach to ensure consistency.

Database Selection: Amazon DynamoDB for Consistency and Performance

• Evaluated various database options, including SQL Server, Amazon Aurora, and Amazon DynamoDB.
• Chose Amazon DynamoDB due to its ability to provide the required high availability, predictable performance, and ACID (Atomicity, Consistency, Isolation, Durability) transactions.
• Designed the DynamoDB table schema to leverage partition keys and sort keys to achieve idempotency and isolation for account balance updates.
• Leveraged DynamoDB Transactions API to perform atomic, consistent, and durable updates to account balances.
• Achieved high performance, reaching 1,200 transactions per second on average with 79ms latency for large accounts.

Migration Strategy: Gradual Dark Launch and Validation

• Implemented a dark launch strategy, where the new architecture processes all transactions in parallel with the existing mainframe-based system.
• In the shadow traffic mode, the new architecture validates each transaction against the existing system to ensure correctness before migrating customers.
• Gradually migrated customers from the legacy system to the new cloud-based platform, ensuring a seamless transition.
• Utilized AWS services like Amazon Kafka (MSK) and Amazon SQS to decouple the synchronous authorization flow from the asynchronous dispatch process.

Key Lessons and Takeaways

• Collaboration, experimentation, and a shared purpose are essential for successful transformation.
• Invest in building a strong foundation, understanding how each AWS service fits your use case, and validating performance before going to production.
• Avoid complex solutions when simpler ones can achieve the same goals. Focus on trade-offs in operations, observability, and cost.
• Leverage AWS resources, such as free digital courses and labs, to accelerate your learning and development.
• Customer obsession should be a guiding principle, as demonstrated by the DynamoDB team's responsiveness to Itaú's support case.

Conclusion

Itaú's journey to modernize its mission-critical checking account platform showcases the power of cloud-based architectures, the importance of high availability and data consistency, and the value of a well-planned migration strategy. By leveraging AWS services and adopting a cell-based design, Itaú was able to achieve its goals of scalability, resilience, and customer-centricity, setting the stage for continued innovation and growth.