GraalVM: Database Integration, Serverless Innovation and the Future

Jan 13, 2026

Thomas Wuerthinger, a Runtime and Compiler Engineer at Oracle, dives into GraalVM's intriguing developments. He clarifies the shift to focusing on LTS Java releases while expanding polyglot support, particularly for Python due to its AI rise. Wuerthinger discusses innovative ideas like serverless workloads directly in Oracle Database and user-space application snapshotting for faster restarts. He emphasizes memory savings and efficiency in deployment, touching on the future roadmap that blends database and serverless technologies seamlessly.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

INSIGHT

GraalVM And Oracle Database Synergy

The GraalVM team is now closely aligned with Oracle Database and the Oracle Multilingual Engine (MLE).
This alignment will prioritize enterprise features that run languages like JavaScript and Python inside the database.

INSIGHT

Database As Runtime For Serverless Logic

GraalVM will enable JVM-based stored procedures and shipping temporary lambdas into the database for execution.
Treating the database as an operating system unlocks efficient, highly isolated serverless workloads close to data.

INSIGHT

Snapshotting For Fast Scale-to-Zero

GraalVM is investing in application snapshotting to enable efficient scale-to-zero and fast cold starts without kernel privileges.
Snapshots can store application state in the database and resume execution from that point.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

An airhacks.fm conversation with Thomas Wuerthinger (@thomaswue) about:

clarification of GraalVM release cadence changes and decoupling from openJDK releases, GraalVM focusing on LTS Java releases only (skipping non-LTS like Java 26), GraalVM as a multi-vendor polyglot project with community edition and third-party vendors like Red Hat BellSoft and microdoc, increased focus on python support due to AI popularity, GraalVM team alignment with Oracle Database organization, Oracle Multilingual Engine (MLE) for running JavaScript and Python in Oracle Database, MySQL MLE integration, native image support for stored procedures in Oracle Database, shipping lambda functions from client applications to database for temporary execution, treating Oracle Database as an operating system for running business logic, serverless workloads directly in Oracle Database, application snapshotting similar to CRaC but running in user space without kernel privileges, efficient scale-to-zero capabilities with native images, Oracle REST Data Services service generalization for serverless execution platform, database triggers for workflow systems and application wake-up, durable functions with transactional state storage in Oracle Database, comparison to AS400 architecture with transaction manager database and operating system in same memory, memory price increases making GraalVM native image more attractive, lower memory consumption benefits of native image beyond just startup time, CPU-based inference support with SIMD and Vector API, TornadoVM for GPU-based inference built on Graal compiler, WebAssembly compilation target for native images, edge function deployment with WebAssembly, Intel memory protection keys for sandboxed native image execution, native image layers for shared base libraries similar to docker layers, profile-guided optimizations for size reduction, upx binary compression for 3x size reduction, memory savings from eliminated class metadata and profiling data not garbage collector differences, 32-bit object headers in serial GC smaller than HotSpot, polyglot integration allowing Python and JavaScript embedding in Java applications, Micronaut framework compile-time annotation processing, quarkus framework best alignment with native image for smallest binaries, GraalVM roadmap focused on database synergies and serverless innovation

Thomas Wuerthinger on twitter: @thomaswue