Java 27 Officially Released with G1 as Default Everywhere and Post-Quantum Support
OpenJDK 27 officially released. G1 collector is now default everywhere, with compact object headers and post-quantum key exchange.
Oracle announced the general availability of OpenJDK 27 on September 15, 2026. It marks Java 27 reaching general availability. According to reporting by Michael Larabel of Phoronix, it was revealed in an announcement that morning. This release brings a major change to the default garbage collector settings. The Garbage-First collector, G1, is now the default in all environments.
Previously, since JDK 9, G1 had been the default only in server environments. In constrained environments, the Serial collector was selected. In Java 27, that distinction has been removed. The reason is that G1 is now competitive at all heap sizes. Its balance of throughput and latency, and of footprint and startup time, was evaluated favorably.
Background Behind Standardizing the G1
Collector Everywhere
The rationale for the change is described in JEP 523. The proposal discontinues the automatic selection of Serial in constrained environments. It is premised on improvements to G1 across all metrics. In recent releases, native memory usage has also been reduced. It has been confirmed to have reached a level comparable to Serial.
JEP 522 contributed significantly to improving maximum throughput. By reducing synchronization, it achieved throughput close to that of Serial. G1 has long had an advantage in maximum latency. Its design, which reclaims the old generation through incremental collections, has proven effective. Because it does not depend on full collections, it can keep growth in pause times in check.
The proposal states:
Since then, we have improved the G1 collector across all metrics, and testing shows that G1 is now competitive with Serial at all heap sizes. With our recent work to reduce synchronization (JEP 522), G1’s maximum throughput is close to that of Serial. G1’s maximum latencies have always been better than those of Serial, since G1 reclaims memory in the old generation via incremental garbage collections rather than full collections.
It also mentions reduced native memory usage. There is also an aim to make G1’s behavior easier to understand. Eliminating differences in collectors by environment makes reasoning easier. This will help reduce the burden of investigation and performance tuning in operations. This article is based on Phoronix (All Rights Reserved) and relies on fair quotation under Article 32 of the Japanese Copyright Act.
Serial had an advantage in simplicity for small heaps. It was valued for fast startup and a lightweight implementation. However, optimization of G1 has narrowed the gap. It was judged that G1 can replace Serial in all situations. A long track record in production supported this shift.
Post-Quantum Cryptography and Compact Object
Headers Introduced Together
The collector is not the only major item in Java 27. Support for post-quantum hybrid key exchange for TLS 1.3 has been added. This strengthens communication protection against future decryption threats. It is a measure aimed at adoption in enterprise systems and public infrastructure. It aligns with moves to ensure cryptographic agility.
The enablement of compact object headers by default is also significant. Object headers are reduced from 96 bits to 64 bits. The goal is to improve packing density and locality. It also directly reduces heap usage. In large-scale deployments, memory savings accumulate.
Compact object headers have been tested since JDK 24. Stability and performance have been confirmed, meeting the criteria for standardization. Benefits when combined with compressed references have also been reported. Workloads that create objects at high density benefit most. It contributes to stable latency in microservices and communication infrastructure.
Updates continue on the language and library fronts as well. The latest preview of late constants is included. Structured concurrency has reached its seventh preview. The latest incubator version of the Vector API is provided. Efforts to improve the portability of parallel computation are continuing.
Practical Impact for Developers and
Operations Teams
Unifying the default collector affects the development process. Developers will need to worry less about behavioral differences due to environments. Reproducibility between local testing and production environments will improve. Performance test configurations can also be simplified. Validation items during migration may be reduced.
In operations, monitoring metrics will need to be reviewed. Collection pauses and throughput trends will change. Tuning values premised on Serial will need to be reevaluated. Verification is especially essential in container environments with small memory. Actual measurements of startup time and footprint will inform decisions.
Java follows a six-month release cadence. It is positioned as an interim release, not a long-term support release. Even so, a change to the defaults has wide repercussions. It offers a preview of the behavior of future long-term support releases. Organizations that validate early can spread out the migration burden.
In terms of compatibility, options remain. Serial can still be explicitly specified via a command-line option. Existing launch configurations will not immediately become invalid. However, review in light of the intent behind the default change is advisable. Casually pinning the old default leaves maintenance complexity.
Future of Structured Concurrency and
Language Features
Structured concurrency is a mechanism for structuring the management of concurrent tasks. It ties the lifetime of subtasks to a lexical scope. It clarifies handling of cancellation and error propagation. Combined with virtual threads, it can improve resource efficiency. It is an area related to controlling agentic processing as discussed in LangGraph vs AutoGen vs CrewAI: AI Agent Comparison 2026.
The seventh preview shows continued refinement. Adjustments to the API shape and cancellation semantics have been repeated. Feedback from users will be important for final adoption. Trial use and reports in practice will determine finalization of the specification. The ultimate goal is to reduce misuse of concurrent programming.
Late constants are a mechanism to increase flexibility in class loading. Resolution of constants can be deferred until needed. It helps shorten startup time and improve initialization order. It has high affinity with frameworks and code generation. Evaluation at the preview stage will decide its future direction.
The Vector API aims to leverage single instruction, multiple data operations. It allows direct use of CPU parallel-computation features from Java. Benefits are expected in machine-learning preprocessing and media processing. It remains in the incubator stage, and the specification will continue to evolve. Ensuring performance portability remains a challenge.
Editorial Opinion
On the short-term impact. We expect standardization of G1 everywhere to increase validation work on container platforms over the next three to six months. We expect small images premised on Serial will require remeasurement of startup time and footprint. We expect framework providers to publish baselines with default settings and provide migration guidance. We expect operators to compare the distribution of collection pause times and review alert thresholds. Disruption will be limited, but we see early validation as a competitive advantage.
On the long-term view. In one to three years, we see the combination of memory efficiency and communication protection influencing platform selection. We see compact object headers helping to contain cloud computing costs. We see post-quantum hybrid key exchange being incorporated into procurement requirements. We see the establishment of structured concurrency and virtual threads changing the norms of concurrent design. We assess that Java’s conservative evolution will lead to earned trust.
Questions from the editors. The question remains whether full standardization prioritized simplicity or room for optimization. We see the question of whether G1 is always optimal in constrained environments as something to be verified by actual measurement. We also see a distributional question of who bears the performance cost of post-quantum support. We see that repeated previews of language features cannot be completed without user participation. We assess that reports from practice will determine specification quality.
References
- “Java 27 Reaches GA With The G1 Garbage Collector By Default Everywhere”, by Michael Larabel — Phoronix, 2026-09-15T13:39:00.000Z (ARR)
- Source URL: https://www.phoronix.com/news/OpenJDK-27-Java-27
Frequently Asked Questions
- What changed with G1 collector standardization in Java 27?
- Previously G1 was the default only in server environments, with Serial selected in constrained environments. In Java 27 the distinction is gone and G1 is the default in all environments. The change is based on JEP 523, grounded in improvements to throughput, latency, and footprint.
- What are the benefits of standardizing compact object headers?
- Object headers are reduced from 96 bits to 64 bits. Improved packing density and data locality, and reduced heap usage are expected. Standardized after validation since JDK 24, memory savings accumulate in large-scale deployments.
- What is the significance of post-quantum hybrid key exchange support for TLS 1.3?
- It strengthens communication protection against future decryption threats from quantum computers. It combines conventional key exchange with post-quantum methods to achieve both security and compatibility. It is a measure aimed at long-term operation in enterprise systems and public infrastructure.
Comments