]> 2026-09-08T16:34:10+09:00 Lesson:kangaroo caching billions of tiny objects on flash 90226bd8 3000 ko Large flash caches for social, IoT, and other tiny-object workloads. Limits: Tradeoffs depend on object-size distribution, write budget, DRAM/flash sizing, trace locality, and set contention. Kangaroo combines a large set-associative KSet with a small log-structured KLog that batches objects before rewriting sets. S3ResearchAgent medium Venue: SOSP. Year: 2021. Set-associative caches minimize DRAM but rewrite flash often; log-structured caches amortize writes but need large DRAM indexes. Verification: official_abstract; confidence=high. 2026-07-16T14:57:56Z 2461238.1235648 Kangaroo: Caching Billions of Tiny Objects on Flash. SOSP 2021. A small write-optimized admission/staging tier can make a DRAM-efficient main cache write-efficient too. Kangaroo: Caching Billions of Tiny Objects on Flash workloads=Facebook traces; Twitter traces; production Facebook deployment; baselines=best prior DRAM-optimized flash cache; best prior write-optimized flash cache; metrics=miss ratio; DRAM bits/object; flash writes; results=29% fewer misses than state of the art; Pareto-optimal across evaluated budgets Can a flash cache store billions of roughly 100-byte objects with both tiny DRAM metadata and low flash write amplification? lab Combine complementary cache organizations at unequal sizes to bridge conflicting metadata and write objectives. Draft 2026-07-18T14:58:38Z 2461240.1240509 2026-07-18T14:58:38Z 2461240.1240509 kangaroo caching billions of tiny objects on flash 90226bd8 3000 ev_d0efc79569484690 2026-07-16T14:57:57Z 2461238.1235764 S3ResearchAgent Kangaroo: Caching Billions of Tiny Objects on Flash. SOSP 2021. paper Bibliographic paper record. not_recorded kangaroo caching billions of tiny objects on flash 90226bd8# 40a143d26629be9ce12edf31b91416f3 3000 verified-content-v1-0095 2026-07-16T18:55:01Z 2461238.288206 S3ResearchAgent Sara McAllister et al., "Kangaroo: Caching Billions of Tiny Objects on Flash", SOSP 2021. paper Verification: official_abstract; confidence=high. Canonical title: Kangaroo: Caching Billions of Tiny Objects on Flash Question: Can a flash cache store billions of roughly 100-byte objects with both tiny DRAM metadata and low flash write amplification? Context: Set-associative caches minimize DRAM but rewrite flash often; log-structured caches amortize writes but need large DRAM indexes. Method: Kangaroo combines a large set-associative KSet with a small log-structured KLog that batches objects before rewriting sets. Evaluation: workloads=Facebook traces; Twitter traces; production Facebook deployment; baselines=best prior DRAM-optimized flash cache; best prior write-optimized flash cache; metrics=miss ratio; DRAM bits/object; flash writes; results=29% fewer misses than state of the art; Pareto-optimal across evaluated budgets Interpretation: A small write-optimized admission/staging tier can make a DRAM-efficient main cache write-efficient too. Reusable lesson: Combine complementary cache organizations at unequal sizes to bridge conflicting metadata and write objectives. Applicability: Large flash caches for social, IoT, and other tiny-object workloads. Limits: Tradeoffs depend on object-size distribution, write budget, DRAM/flash sizing, trace locality, and set contention. not_recorded kangaroo caching billions of tiny objects on flash 90226bd8# 34d7b7ec02c9b2ee4d3b91d0d437b118 3000 canonical-paper-v2-90226bd8 2026-07-18T05:25:35Z 2461239.7260995 S3ResearchAgent Sara McAllister et al., "Kangaroo: Caching Billions of Tiny Objects on Flash", SOSP 2021. paper Verification: official_abstract; confidence=medium. Canonical title: Kangaroo: Caching Billions of Tiny Objects on Flash Question: Can a flash cache store billions of roughly 100-byte objects with both tiny DRAM metadata and low flash write amplification? Context: Set-associative caches minimize DRAM but rewrite flash often; log-structured caches amortize writes but need large DRAM indexes. Method: Kangaroo combines a large set-associative KSet with a small log-structured KLog that batches objects before rewriting sets. Evaluation: workloads=Facebook traces; Twitter traces; production Facebook deployment; baselines=best prior DRAM-optimized flash cache; best prior write-optimized flash cache; metrics=miss ratio; DRAM bits/object; flash writes; results=29% fewer misses than state of the art; Pareto-optimal across evaluated budgets Interpretation: A small write-optimized admission/staging tier can make a DRAM-efficient main cache write-efficient too. Reusable lesson: Combine complementary cache organizations at unequal sizes to bridge conflicting metadata and write objectives. Applicability: Large flash caches for social, IoT, and other tiny-object workloads. Limits: Tradeoffs depend on object-size distribution, write budget, DRAM/flash sizing, trace locality, and set contention. official_abstract kangaroo caching billions of tiny objects on flash 90226bd8# 80cf09d53473c03bc5acb4485547d2ea 3000 2026-07-18T14:58:37Z 2461240.1240394 S3ResearchAgent verify_ded4c709ddb702758382 partial_source context 보존 원문 객체를 확보했으나 이 일괄 검증에서는 claim-bearing 범위를 재판정하지 않아 결론을 보류함. inconclusive 945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 보존 파일 objects/sha256/94/945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 ev_d0efc79569484690 402cc5150376ab710bed45fc56357b9023568a74c5ed68dcdb397a4c3c822a65 kangaroo caching billions of tiny objects on flash 90226bd8# 11e808e45026411522a271ed92df4367 3000 2026-07-18T14:58:37Z 2461240.1240394 S3ResearchAgent verify_fc536f7f6cf48e60bc0d partial_source context 보존 원문 객체를 확보했으나 이 일괄 검증에서는 claim-bearing 범위를 재판정하지 않아 결론을 보류함. inconclusive 945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 보존 파일 objects/sha256/94/945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 verified-content-v1-0095 b2237eab506082f054976b052806764d5bfff9c865624cb581dd6f4f057761c6 kangaroo caching billions of tiny objects on flash 90226bd8# 0733c8c8fb28ba8020a855d96fbb343c 3000 2026-07-18T14:58:38Z 2461240.1240509 S3ResearchAgent verify_d832c7283d7b19d7161a partial_source context 보존 원문 객체를 확보했으나 이 일괄 검증에서는 claim-bearing 범위를 재판정하지 않아 결론을 보류함. inconclusive 945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 R2-RESTIC:7f893ca5afd2cfb6fe320e9b61063ccc70e75a7a96589420038c8cf338b273be; archive-manifest-sha256=e28171fb69e141ce306d92dfe4b10e6cdc6e81d4fa910c30a846204dbcf8edf8; sha256=945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 보존 파일 objects/sha256/94/945b0133c4f9661ab5408194737209d8f17d5fac0b138f4c873dbab4f779ced0 canonical-paper-v2-90226bd8 f66afb0eb548cd5681be6973107fde749842f44e3ba3cfe4bc927f4f093ed3a0 kangaroo caching billions of tiny objects on flash 90226bd8# 29ea2237e383ecbb5f9d8d6034ea32c4