Search before reporting

• I searched in the issues and found nothing similar.

Read release policy

• I understand that unsupported versions don't get bug fixes. I will attempt to reproduce the issue on a supported version of Pulsar client and Pulsar broker.

User environment

master branch version of Pulsar, 34f8657

Issue Description

There's a opportunity to significantly reduce memory allocations in Pulsar by caching parsed message metadata. In a simple profiling test case in #24623 I noticed that
there's a lot of allocations coming from org.apache.pulsar.common.protocol.Commands.peekAndCopyMessageMetadata.

The broker cache can be modified to support caching the message metadata together with the entries so that unnecessary parsing and allocations can be avoided.

Error messages

Reproducing the issue

With #24623 changes, enabling the BrokerEntryCacheMultiBrokerTest.testTailingReadsRollingRestart test
by modifying the @Test(enabled = false) annotation to @Test(invocationCount = 5) and then running

• Download Async Profiler from https://github.com/async-profiler/async-profiler/releases and after installing set LIBASYNCPROFILER_PATH environment variable to the full path of the libasyncProfiler.so (Linux) or libasyncProfiler.dylib (MacOS) file.
• You can use Eclipse Mission Control (https://adoptium.net/jmc) to open the JFR file and analyze it.

export LIBASYNCPROFILER_PATH=/full/path/to/async-profiler/lib/libasyncProfiler.so (or .dylib on MacOS)
mvn -PtestAsyncProfiler test -pl pulsar-broker -Dtest=BrokerEntryCacheMultiBrokerTest

Additional information

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!