속성으로 검색

이 문서는 속성과 이름이 지정된 값으로 설명된 개체를 찾기 위한 간단한 탐색 인터페이스를 제공합니다. 그 밖에 사용 가능한 검색 인터페이스에는 문서 속성 검색 및 Ask 쿼리 빌더가 있습니다.

결과 목록

• Lesson:technical review dram 0a5cae63  + (L2P 세그먼트와 유효성 비트맵을 함께 지역화하면 작은 SRAM에서도 별도 전역 비트맵의 flash metadata I/O를 피할 수 있다. 다만 메모리 오버헤드와 write amplification trade-off는 공식 초록으로 확인하지 못했다.)
• Lesson:dram 8d56adf1  + (L2P 세그먼트와 유효성 비트맵을 함께 지역화하면 작은 SRAM에서도 별도 전역 비트맵의 flash metadata I/O를 피할 수 있다. 다만 메모리 오버헤드와 write amplification trade-off는 공식 초록으로 확인하지 못했다.)
• Lesson:corobase coroutine oriented main memory database engine c4a28018  + (Language/runtime continuations can hide latency while preserving readable transactional control flow.)
• Lesson:technical review transparent and selective real time interrupt services for performance improvem b78f264b  + (Linux 2.6와 Hackbench 중심 결과이고, interrupt postponement가 허용되는 device/workload 범위를 다른 시스템에 그대로 일반화할 수 없다. Technical interpretation: 모든 interrupt를 동일한 real-time 경로로 처리하지 말고 deadline-criticality와의 관련성을 기준으로 분리해야 한다.)
• Lesson:transparent and selective real time interrupt services for performance improvement 6b173779  + (Linux 2.6와 Hackbench 중심 결과이고, interrupt postponement가 허용되는 device/workload 범위를 다른 시스템에 그대로 일반화할 수 없다. Technical interpretation: 모든 interrupt를 동일한 real-time 경로로 처리하지 말고 deadline-criticality와의 관련성을 기준으로 분리해야 한다.)
• Lesson:pivot-b200-20260717-tmux-argument-boundary-a01  + (Local quoting did not survive SSH's remote command reconstruction as a single tmux argument. tmux then concatenated multiple send-keys string arguments without inserting spaces.)
• Lesson:pet proactive demotion for efficient tiered memory management e2185da3  + (Logical allocation boundaries can be a better management unit than fixed VM pages.)
• Lesson:machine validation receipts should be structurally unable to grant human acceptance 470a7c00  + (Machine validation establishes artifact in … Machine validation establishes artifact integrity and an inventory binding, while normal Git/PR review records human accountability and final-run authorization stays outside the validator. Historical V1 receipt strings are immutable evidence, not an active attestation gate. evidence, not an active attestation gate.)
• Lesson:ozz identifying kernel out of order concurrency bugs with in vivo memory access reordering 0b85f639  + (Making weak-memory effects controllable turns a probabilistic hardware behavior into a testable schedule dimension.)
• Lesson:predicting future system reliability with a component level dram fault model 1f248bb6  + (Mapping observed errors to physical components enables architecture-specific projections instead of reusing aggregate failure rates.)
• Lesson:fast on device llm inference with npus deecd917  + (Mobile heterogeneous execution can turn an otherwise rigid NPU into the dominant prefill engine.)
• Lesson:serverlessllm low latency serverless inference for large language models 69b8cd6d  + (Model loading must be treated as a scheduled data-placement pipeline rather than a one-off cold-start operation.)
• Lesson:cacheblend fast large language model serving for rag with cached knowledge fusion 056a5f5e  + (Most cached chunk states remain usable if a small, well-chosen subset repairs cross-context interactions.)
• Lesson:an empirical study on low gpu utilization of deep learning jobs c5389b31  + (Most low utilization is an actionable software-pipeline problem rather than an immutable GPU limit.)
• Lesson:tierscape harnessing multiple compressed tiers to tame server memory tco 9ed9ed0d  + (Multiple configurable tiers offer a richer placement frontier than a fixed two-tier design.)
• Lesson:scalable billion point approximate nearest neighbor search using smartssds 3c4169e5  + (Near-data ANN needs coordinated indexing, pruning, and scheduling, not simple operator offload.)
• Lesson:aegonkv a high bandwidth low tail latency and low storage cost kv separated lsm store with smart 5a081f42  + (Near-data GC can isolate maintenance bandwidth and CPU from the foreground path.)
• Lesson:m5 mastering page migration and memory management for cxl based tiered memory systems 4e752168  + (Near-memory visibility supports finer placement decisions without heavy CPU profiling.)
• Lesson:technical review daac device reserved memory as an eviction based file cache f0f149f0  + (Nexus S 세대의 Android/flash workload가 평가 범위다. eCache는 일반 anonymous page가 아니라 page-cache victim을 저장한다. Technical interpretation: 빨리 회수해야 하는 borrowed memory에는 재구성 가능한 clean cache data를 우선 배치해야 한다.)
• Lesson:daac device reserved memory as an eviction based file cache db4ed198  + (Nexus S 세대의 Android/flash workload가 평가 범위다. eCache는 일반 anonymous page가 아니라 page-cache victim을 저장한다. Technical interpretation: 빨리 회수해야 하는 borrowed memory에는 재구성 가능한 clean cache data를 우선 배치해야 한다.)
• Lesson:technical review enlightening the i o path a holistic approach for application performance dabf0f9c  + (Non-critical I/O를 하나로 제한하면 foreground late … Non-critical I/O를 하나로 제한하면 foreground latency는 보호하지만 system throughput을 낮춘다. 제한 완화 시 throughput은 223에서 345MB/s로 늘지만 foreground 평균 latency는 110에서 1771마이크로초로 악화됐다. User-level fiber dependency를 직접 검출하지 못하며 multi-tenant fairness와 replicated store는 미래 작업이다.</br></br>Technical interpretation: Priority는 I/O 발행 시점의 정적 속성이 아니라 request dependency를 따라 계층 전체에서 전파되어야 한다.라 request dependency를 따라 계층 전체에서 전파되어야 한다.)
• Lesson:enlightening the i o path a holistic approach for application performance a3c2b020  + (Non-critical I/O를 하나로 제한하면 foreground late … Non-critical I/O를 하나로 제한하면 foreground latency는 보호하지만 system throughput을 낮춘다. 제한 완화 시 throughput은 223에서 345MB/s로 늘지만 foreground 평균 latency는 110에서 1771마이크로초로 악화됐다. User-level fiber dependency를 직접 검출하지 못하며 multi-tenant fairness와 replicated store는 미래 작업이다.</br></br>Technical interpretation: Priority는 I/O 발행 시점의 정적 속성이 아니라 request dependency를 따라 계층 전체에서 전파되어야 한다.라 request dependency를 따라 계층 전체에서 전파되어야 한다.)
• Lesson:pact a criticality first design for tiered memory 20db3445  + (Online page placement should optimize attributed CPU stall time rather than access count.)
• Lesson:mitigating application resource overload with targeted task cancellation 90270bf8  + (Overload control improves when it attributes pressure to active culprit tasks and invokes application-safe cancellation, rather than rejecting arriving victim requests or statically partitioning resources.)
• Lesson:scalable address spaces using concurrent interval skiplist d1f1b9c0  + (Parallel interval operations require one data structure that jointly defines lookup and conflict boundaries.)

• Lesson:tiered memory management beyond hotness 6ad152df  + (Placement should optimize exposed stall cost, not raw access counts.)

• Lesson:how storage i o affects user perceived latency in mobile apps b59ccb75  + (Poster 목록은 발표의 존재와 서지정보만 뒷받침한다. 제목만으로 storage I/O가 체감 지연의 병목이었다거나 특정 최적화가 효과적이었다고 결론 내릴 수 없다.)
• Lesson:ice collaborating memory and process management for user experience on resource limited mobile d 312e453f  + (Process lifecycle control can prevent memory-reclaim feedback loops that page replacement alone cannot resolve.)
• Lesson:kvcache cache in the wild characterizing and optimizing kvcache cache at a large cloud provider 0b39b53c  + (Production KV-cache reuse는 multi-turn 여부 하나로 설명되지 않습니다. To-B API workload에서는 반복 system prompt 때문에 single-turn 요청이 reuse를 지배할 수 있고, request category별 시간/공간 locality와 짧은 lifespan을 함께 모델링해야 cache size와 eviction policy를 정할 수 있습니다.)
• Lesson:flexmem adaptive page profiling and migration for tiered memory 43991e2a  + (Profiling fidelity and migration aggressiveness should be controlled together from workload feedback.)
• Lesson:technical review request aware cooperative i o scheduling for scale out database applications fe4a3e09  + (Prototype은 read-oriented scan만 구현했다. Write … Prototype은 read-oriented scan만 구현했다. Write merging, group commit, journaling에서 request ID가 섞이는 문제, background I/O의 request-context 귀속, replication은 미해결이다.</br></br>Technical interpretation: 분산 요청의 tail을 줄이려면 각 노드의 local fairness보다 원래 사용자 요청의 global order를 I/O 계층까지 보존하는 것이 중요할 수 있다.의 global order를 I/O 계층까지 보존하는 것이 중요할 수 있다.)
• Lesson:request aware cooperative i o scheduling for scale out database applications 25327d2c  + (Prototype은 read-oriented scan만 구현했다. Write … Prototype은 read-oriented scan만 구현했다. Write merging, group commit, journaling에서 request ID가 섞이는 문제, background I/O의 request-context 귀속, replication은 미해결이다.</br></br>Technical interpretation: 분산 요청의 tail을 줄이려면 각 노드의 local fairness보다 원래 사용자 요청의 global order를 I/O 계층까지 보존하는 것이 중요할 수 있다.의 global order를 I/O 계층까지 보존하는 것이 중요할 수 있다.)
• Lesson:disaggregated raid storage in modern datacenters 4eb679fe  + (RAID control/data paths should exploit the network topology and overlap protection work rather than emulate a local controller.)
• Lesson:pacar improved buffered i o locality on numa systems with page cache replication adc0c538  + (Read-mostly page-cache replication can trade controlled coherence for substantial NUMA locality.)
• Lesson:the cost of dynamic reasoning demystifying ai agents and test time scaling from an ai infrastruc 6d6dff58  + (Reasoning quality must be optimized jointly with tail latency, energy, and capacity rather than token count alone.)
• Lesson:pmr fast application response via parallel memory reclaim on mobile devices 4c59e491  + (Reclamation latency falls when dirty work is predicted, separated, and shaped for the storage device.)
• Lesson:hotrap hot record retention and promotion for lsm trees with tiered storage 849aa350  + (Record-level popularity can guide selective LSM reshaping more efficiently than run-level movement.)
• Lesson:polymorphic error correction 25fcf9e1  + (Redundancy can be encoded for multiple operating points instead of hard-wiring one reliability/security tradeoff.)
• Lesson:usher holistic interference avoidance for resource optimized ml inference 74231fc1  + (Resource estimation, placement, and graph-level execution must be coordinated to make GPU multiplexing economical.)
• Lesson:citadel rethinking memory allocation to safeguard against inter domain rowhammer exploits 8fd39aca  + (Row-aware allocator structure can turn physical separation into a scalable isolation primitive.)
• Lesson:understanding silent data corruptions in a large production cpu population 56bebc4c  + (SDC policy must distinguish stable, testable faults from intermittent environment-dependent faults.)
• Lesson:extending applications safely and efficiently 7a5f9f65  + (Safe extensibility needs a contract covering both what code may access and what resources it may consume.)
• Lesson:fast flexible and practical kernel extensions 9eb8897f  + (Safety can be decomposed by ownership, allowing stronger guarantees for kernel state without proving every extension invariant statically.)
• Lesson:ic cache efficient large language model serving via in context caching 4f05d8b7  + (Semantic history can be an executable cache that substitutes cheaper conditioned inference for repeated full reasoning.)
• Lesson:mitigating resource usage dependency in sorting based kv stores on hybrid storage devices via op 31cbb8ad  + (Separating lightweight index merging from value placement removes avoidable data rewriting.)
• Lesson:constructing and analyzing the lsm compaction design space 1aa4ace5  + (Separating policy primitives reveals which choice causes each tradeoff and enables deliberate hybrid designs.)
• Lesson:emt an os framework for new memory translation architectures 6d3356f6  + (Separating translation semantics from page-table implementation makes architecture research deployable in a full OS.)
• Lesson:technical review d2fq device direct fair queueing for nvme ssds c60d73e1  + (Short-term unfairness의 이론적 상한은 없고 실험적으로만 확인했다. NVMe WRR 지원 장치가 필요하며 주 평가는 Samsung Z-SSD에 의존한다. 장치 queue 수 제약도 있다.)
• Lesson:d2fq device direct fair queueing for nvme ssds 36293af6  + (Short-term unfairness의 이론적 상한은 없고 실험적으로만 확인했다. NVMe WRR 지원 장치가 필요하며 주 평가는 Samsung Z-SSD에 의존한다. 장치 queue 수 제약도 있다.)
• Lesson:scalable far memory balancing faults and evictions edcf1536  + (Slightly less precise eviction can be worthwhile when synchronization otherwise destroys scalability.)
• Lesson:fast trapless kernel probes everywhere 781e739b  + (Small compile-time layout support removes a runtime observability tax while preserving compatibility.)