26.08.0

Optimizer Hub now supports Compilation Streaming. Optimizer Hub builds a stream of previously performed compilations from its Code Cache and the promoted ReadyNow profile, and serves that stream to JVMs at startup. A JVM installs the streamed compilations directly instead of requesting them one at a time as the application warms up, which shortens the time needed to clear the compilation queue and reduces load on Cloud Native Compiler.

Compilation Streaming requires Azul Zing 26.08.1.0 or later, and must be enabled on both the server and the JVM. On the server, set compilationStreaming.enabled. On the JVM, add -XX:+CNCEnableCompilationStreaming. See Installing Optimizer Hub with Compilation Streaming for the complete commands, and Using Compilation Streaming for the full reference.

The task-executor component is now deployed by default. It builds the compilation streams used by Compilation Streaming, and refreshes the ReadyNow profiles used by Indirect Delivery, so neither feature needs a separate opt-in anymore. The pod is sized at 4 CPU and 16 GiB of memory, with 10 GiB of ephemeral storage for temporary stream files. Like gw-proxy, mgmt-gateway, and the operator, task-executor is a fixed-size component that does not autoscale, and its cores are now included in the simpleSizing vCore budget. If you use neither Compilation Streaming nor Indirect Delivery, you can disable it with taskExecutor.enabled: false.

Optimizer Hub now supports Indirect Delivery of ReadyNow profiles. Instead of sending the input profile to the JVM over gRPC, Optimizer Hub stores the profile in blob storage and hands the JVM a pre-signed URI, which the JVM uses to download the profile over HTTP. Use this to reduce the volume of traffic through the Optimizer Hub gateway, or when your network topology makes a direct gRPC transfer of large profiles impractical. Enable it on the server with indirectDelivery.enabled: true. See Indirect Delivery of ReadyNow Profiles.

In Azul Zing Builds of OpenJDK, version 26.08.0.0 and newer, methods compiled at a reduced optimization level during a local fallback period, as capped by -XX:CNCLocalFallbackOptLevelLimit, are guaranteed to be recompiled at their intended level once Cloud Native Compiler is reachable again. See Automatic Switch Back to Cloud Native Compiler.

Includes bug fixes for Optimizer Hub 26.05.0.