TornadoVM Collaboration and GPU Strategy

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