An overview of the demo application using DynamoDB, Aurora DSQL with JDBC, and Hibernate to set the stage for performance comparisons.
A review of the data access object (DAO) implementations for DynamoDB, plain JDBC with HikariPool, and the Hibernate ORM framework.
The Lambda execution lifecycle is explained to differentiate cold starts from warm starts, followed by baseline performance measurements for each data access pattern.
AWS Lambda SnapStart improves startup times by creating a pre-initialized snapshot, which can be further enhanced using priming techniques with CRaC hooks.
Performance benchmarks show significant cold start reduction from priming, and the behavior of the tiered, low-latency cache explains why later invocations are faster.
GraalVM's ahead-of-time compilation offers peak performance but requires extensive configuration for reflection and resources, especially when using frameworks like Hibernate.
A direct performance comparison shows GraalVM outperforming SnapStart, alongside a brief look at related technologies like Lambda managed instances and Project Leyden.
Choose SnapStart for a managed solution with good performance via priming, or select GraalVM for maximum performance if you can manage the build complexity.
Jobs that call for the skills explored in this talk.