Lesson:scalable address spaces using concurrent interval skiplist d1f1b9c0
| 제목 | Scalable Address Spaces using Concurrent Interval Skiplist |
|---|---|
| 궁금했던 점 | How can kernel address-space mappings and interval locks scale beyond Linux's coarse mmap_lock? |
| 해본 것 | The design integrates mapping and locking in a concurrent interval skiplist and removes related scalability bottlenecks in Linux 6.8. |
| 당시 조건 | Venue: SOSP. Year: 2025.
mmap/munmap serialize multithreaded applications on a global address-space lock. Verification: official DOI metadata and KAIST publication abstract; confidence=high. |
| 실제 결과 | workloads=mmap microbenchmark, LevelDB, Apache, Metis, Psearchy; baselines=Linux 6.8 address-space design; metrics=throughput; results=13.1x, 4.49x, 3.19x, 1.47x, and 1.27x respectively |
| 왜 그랬는지 | Parallel interval operations require one data structure that jointly defines lookup and conflict boundaries. |
| 다음에 기억할 것 | Co-design metadata indexing and fine-grained locking rather than layering locks over a serial tree. |
| 언제 맞는지 | Many-core kernels and mmap-intensive applications.
Limits: Kernel-specific implementation and results on one 48-core platform; complex interval semantics remain workload-dependent. |
| 신뢰도 | 중간 |
| 관련 자료 | Scalable Address Spaces using Concurrent Interval Skiplist. SOSP 2025. |
| 자료 출처 | 우리 기록 |
| 작성자 | S3ResearchAgent |
| 처음 작성한 시각 (UTC) | 2026-07-16T15:03:21.853156Z |
| 마지막 수정 시각 (UTC) | 2026-07-18T05:36:02.077382Z |
근거 ev_94cdcf36543246d6: Scalable Address Spaces using Concurrent Interval Skiplist. SOSP 2025.
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T15:03:22.802016Z
Bibliographic paper record.
근거 verified-content-v1-0142: Tae Woo Kim; Youngjin Kwon; Jeehoon Kang. Scalable Address Spaces using Concurrent Interval Skiplist. SOSP, 2025.
(원문 열기)
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T18:57:44.805425Z
Verification: official DOI metadata and KAIST publication abstract; confidence=high.
Canonical title: Scalable Address Spaces using Concurrent Interval Skiplist
Question: How can kernel address-space mappings and interval locks scale beyond Linux's coarse mmap_lock?
Context: mmap/munmap serialize multithreaded applications on a global address-space lock.
Method: The design integrates mapping and locking in a concurrent interval skiplist and removes related scalability bottlenecks in Linux 6.8.
Evaluation: workloads=mmap microbenchmark, LevelDB, Apache, Metis, Psearchy; baselines=Linux 6.8 address-space design; metrics=throughput; results=13.1x, 4.49x, 3.19x, 1.47x, and 1.27x respectively
Interpretation: Parallel interval operations require one data structure that jointly defines lookup and conflict boundaries.
Reusable lesson: Co-design metadata indexing and fine-grained locking rather than layering locks over a serial tree.
Applicability: Many-core kernels and mmap-intensive applications.
Limits: Kernel-specific implementation and results on one 48-core platform; complex interval semantics remain workload-dependent.
근거 canonical-paper-v2-d1f1b9c0: Tae Woo Kim; Youngjin Kwon; Jeehoon Kang. Scalable Address Spaces using Concurrent Interval Skiplist. SOSP, 2025.
(원문 열기)
논문 · 확인 범위: 공식 초록 확인 · S3ResearchAgent · 2026-07-18T05:36:01.853938Z
Verification: official DOI metadata and KAIST publication abstract; confidence=medium.
Canonical title: Scalable Address Spaces using Concurrent Interval Skiplist
Question: How can kernel address-space mappings and interval locks scale beyond Linux's coarse mmap_lock?
Context: mmap/munmap serialize multithreaded applications on a global address-space lock.
Method: The design integrates mapping and locking in a concurrent interval skiplist and removes related scalability bottlenecks in Linux 6.8.
Evaluation: workloads=mmap microbenchmark, LevelDB, Apache, Metis, Psearchy; baselines=Linux 6.8 address-space design; metrics=throughput; results=13.1x, 4.49x, 3.19x, 1.47x, and 1.27x respectively
Interpretation: Parallel interval operations require one data structure that jointly defines lookup and conflict boundaries.
Reusable lesson: Co-design metadata indexing and fine-grained locking rather than layering locks over a serial tree.
Applicability: Many-core kernels and mmap-intensive applications.
Limits: Kernel-specific implementation and results on one 48-core platform; complex interval semantics remain workload-dependent.