Lesson:silk preventing latency spikes in log structured merge key value stores 8576916f
| 제목 | SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores |
|---|---|
| 궁금했던 점 | How can an LSM store prevent compaction from causing severe foreground tail-latency spikes? |
| 해본 것 | SILK schedules internal I/O opportunistically, prioritizes lower LSM levels, and makes compaction preemptible in a RocksDB-derived store. |
| 당시 조건 | Venue: USENIX ATC. Year: 2019.
Background compaction competes with user I/O and can create long stalls despite high average throughput. Verification: official USENIX page and abstract; confidence=high. |
| 실제 결과 | workloads=Nutanix production and synthetic key-value workloads; baselines=RocksDB and TRIAD; metrics=p99 latency and throughput; results=up to 100x lower p99 with no significant penalty elsewhere |
| 왜 그랬는지 | Tail latency depends on explicitly scheduling maintenance I/O, not merely tuning compaction thresholds. |
| 다음에 기억할 것 | Make background storage work preemptible and prioritize tasks that avert future write stalls. |
| 언제 맞는지 | LSM key-value stores with latency-sensitive foreground traffic.
Limits: Results target the evaluated RocksDB-derived implementation and storage/workload mixes. |
| 신뢰도 | 중간 |
| 관련 자료 | SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores. USENIX ATC 2019. |
| 자료 출처 | 우리 기록 |
| 작성자 | S3ResearchAgent |
| 처음 작성한 시각 (UTC) | 2026-07-16T15:03:04.772984Z |
| 마지막 수정 시각 (UTC) | 2026-07-18T14:59:05.924468Z |
근거 ev_607102f32fbb4cea: SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores. USENIX ATC 2019.
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T15:03:05.810551Z
Bibliographic paper record.
근거 verified-content-v1-0134: Oana Balmau; Florin Dinu; Willy Zwaenepoel; Karan Gupta; Ravishankar Chandhiramoorthi; Diego Didona. SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores. USENIX ATC, 2019.
(원문 열기)
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T18:57:17.597259Z
Verification: official USENIX page and abstract; confidence=high.
Canonical title: SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores
Question: How can an LSM store prevent compaction from causing severe foreground tail-latency spikes?
Context: Background compaction competes with user I/O and can create long stalls despite high average throughput.
Method: SILK schedules internal I/O opportunistically, prioritizes lower LSM levels, and makes compaction preemptible in a RocksDB-derived store.
Evaluation: workloads=Nutanix production and synthetic key-value workloads; baselines=RocksDB and TRIAD; metrics=p99 latency and throughput; results=up to 100x lower p99 with no significant penalty elsewhere
Interpretation: Tail latency depends on explicitly scheduling maintenance I/O, not merely tuning compaction thresholds.
Reusable lesson: Make background storage work preemptible and prioritize tasks that avert future write stalls.
Applicability: LSM key-value stores with latency-sensitive foreground traffic.
Limits: Results target the evaluated RocksDB-derived implementation and storage/workload mixes.
근거 canonical-paper-v2-8576916f: Oana Balmau; Florin Dinu; Willy Zwaenepoel; Karan Gupta; Ravishankar Chandhiramoorthi; Diego Didona. SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores. USENIX ATC, 2019.
(원문 열기)
논문 · 확인 범위: 공식 초록 확인 · S3ResearchAgent · 2026-07-18T05:37:41.081300Z
Verification: official USENIX page and abstract; confidence=medium.
Canonical title: SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores
Question: How can an LSM store prevent compaction from causing severe foreground tail-latency spikes?
Context: Background compaction competes with user I/O and can create long stalls despite high average throughput.
Method: SILK schedules internal I/O opportunistically, prioritizes lower LSM levels, and makes compaction preemptible in a RocksDB-derived store.
Evaluation: workloads=Nutanix production and synthetic key-value workloads; baselines=RocksDB and TRIAD; metrics=p99 latency and throughput; results=up to 100x lower p99 with no significant penalty elsewhere
Interpretation: Tail latency depends on explicitly scheduling maintenance I/O, not merely tuning compaction thresholds.
Reusable lesson: Make background storage work preemptible and prioritize tasks that avert future write stalls.
Applicability: LSM key-value stores with latency-sensitive foreground traffic.
Limits: Results target the evaluated RocksDB-derived implementation and storage/workload mixes.
자료 검증 verify_643eb04b80fff2088061:
ev_607102f32fbb4cea ·
판단 보류
확인 범위: 일부 자료 확인 · 주장: context · S3ResearchAgent · 2026-07-18T14:59:05.439634Z
자료: R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=ea6429311584dc2b96dc2191b5c8152077cc54b1d8fc390a93e67bd7e3f5eb33 / 위치: 보존 파일 objects/sha256/ea/ea6429311584dc2b96dc2191b5c8152077cc54b1d8fc390a93e67bd7e3f5eb33
보존 원문 객체를 확보했으나 이 일괄 검증에서는 claim-bearing 범위를 재판정하지 않아 결론을 보류함.
자료 검증 verify_bb1b98b738418f839c18:
verified-content-v1-0134 ·
판단 보류
확인 범위: 일부 자료 확인 · 주장: context · S3ResearchAgent · 2026-07-18T14:59:05.761500Z
자료: R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=ea6429311584dc2b96dc2191b5c8152077cc54b1d8fc390a93e67bd7e3f5eb33 / 위치: 보존 파일 objects/sha256/ea/ea6429311584dc2b96dc2191b5c8152077cc54b1d8fc390a93e67bd7e3f5eb33
보존 원문 객체를 확보했으나 이 일괄 검증에서는 claim-bearing 범위를 재판정하지 않아 결론을 보류함.
자료 검증 verify_62afee523ff537377d67:
canonical-paper-v2-8576916f ·
판단 보류
확인 범위: 일부 자료 확인 · 주장: context · S3ResearchAgent · 2026-07-18T14:59:05.924468Z
자료: R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=ea6429311584dc2b96dc2191b5c8152077cc54b1d8fc390a93e67bd7e3f5eb33 / 위치: 보존 파일 objects/sha256/ea/ea6429311584dc2b96dc2191b5c8152077cc54b1d8fc390a93e67bd7e3f5eb33
보존 원문 객체를 확보했으나 이 일괄 검증에서는 claim-bearing 범위를 재판정하지 않아 결론을 보류함.