FAST '21
as the organisers put it
The 19th USENIX Conference on File and Storage Technologies (FAST '21) will take place on February 23–25, 2021, as a virtual event. FAST brings together storage-system researchers and practitioners to explore new directions in the design, implementation, evaluation, and deployment of storage systems.
| title | author | length |
|---|---|---|
| ROART: Range-query Optimized Persistent ART | Shaonan Ma, Kang Chen, Shimin Chen, Mengxing Liu, Jianglang Zhu, Hongbo Kang, Yongwei Wu | 12m |
ROART is a persistent index built on adaptive radix tree that implements leaf compaction to store up to 64 leaves in a contiguous array, reducing pointer… | ||
| SpanDB: A Fast, Cost-Effective LSM-tree Based KV Store on Hybrid Storage | Hao Chen, Chaoyi Ruan, Cheng Li, Xiaosong Ma, Yinlong Xu | 12m |
SpanDB adapts RocksDB to put write-ahead logs and the top LSM levels on a small NVMe while keeping bulk SSTables on cheaper SSDs, implements SPDK for polling… | ||
| Evolution of Development Priorities in Key-value Stores Serving Large-scale Applications | Siying Dong, Andrew Kryczka, Yanqin Jin, Michael Stumm | 11m |
RocksDB evolves over eight years (forked from LevelDB 2012, open‑sourced 2013, production by 2015–16) and shifts optimization goals from reducing write… | ||
| REMIX: Efficient Range Query for LSM-trees | Wenshao Zhong, Chen Chen, Xingbo Wu, Song Jiang | 12m |
Presents REMIX, a space-efficient KV index that records a global sorted view across multiple LSM table files so range queries locate a start via one binary… | ||
| High Velocity Kernel File Systems with Bento | Samantha Miller, Kaiyuan Zhang, Mengqi Chen, Ryan Jennings, Ang Chen, Danyang Zhuo, Thomas E. Anderson | 11m |
Bento is a Linux kernel framework that runs file systems written in safe Rust inside a kernel module while exposing the same interfaces for user‑space… | ||
| Scalable Persistent Memory File System with Kernel-Userspace Collaboration | Youmin Chen, Youyou Lu, Bohong Zhu, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau, Jiwu Shu | 13m |
Kuco is a client–server PM filesystem architecture that maps persistent memory into user space while keeping a tiny kernel KFS to enforce permissions via… | ||
| Rethinking File Mapping for Persistent Memory | Ian Neal, Gefei Zuo, Eric Shiple, Tanvir Ahmed Khan, Youngjin Kwon, Simon Peter, Baris Kasikci | 12m |
Finds that file mapping consumes as much as 70% of the IO path on Intel Optane DC persistent memory and shows page-cache metadata caching hurts performance… | ||
| pFSCK: Accelerating File System Checking and Repair for Modern Storage | David Domingo, Sudarsun Kannan | 12m |
pFSCK parallelizes checking down to inode and directory-block granularity and combines data parallelism with pipeline parallelism plus per-thread I/O caches… | ||
| Pattern-Guided File Compression with User-Experience Enhancement for Log-Structured... | Cheng Ji, Li-Pin Chang, Riwei Pan, Chao Wu, Congming Gao, Liang Shi, Tei-Wei Kuo, Chun Jason Xue | 12m |
FPC is a concrete file-system patch for F2FS implemented on an AOSP embedded board that adds dual-mode compression with foreground/background modes: foreground… | ||
| Keynote Address: Netflix: Streaming Entertainment to 200 Million Members Around the World | Jonathan Looney | 45m |
Netflix runs Open Connect OCAs that stream to ~200 million members and peak above 100 Tbps; caches run FreeBSD on commodity hardware and aim for… | ||
| ArchTM: Architecture-Aware, High Performance Transaction for Persistent Memory | Kai Wu, Jie Ren, Ivy Peng, Dong Li | 14m |
A transaction system tuned to Intel DCPMM hardware, ArchTM uses a copy-on-write variant with a scalable DRAM lookup table to avoid metadata writes, an… | ||
| SPHT: Scalable Persistent Hardware Transactions | Daniel Castro, Alexandro Baldassin, João Barreto, Paolo Romano | 12m |
Paper identifies scalability bottlenecks of prior persistent transactional memory designs by measuring on Intel Optane DC PM rather than emulation, proposes… | ||
| The Dilemma between Deduplication and Locality: Can Both be Achieved? | Xiangyu Zou, Jingsong Yuan, Philip Shilane, Wen Xia, Haijun Zhang, Xuan Wang | 12m |
MFDedup implements a classification-based dedupe plus an offline iterative reorganization that detects duplicates mainly against the previous snapshot and… | ||
| Remap-SSD: Safely and Efficiently Exploiting SSD Address Remapping to Eliminate... | You Zhou, Qiulin Wu, fei wu, Hong Jiang, Jian Zhou, ChangSheng Xie | 13m |
Remap-SSD implements a remap primitive plus a log-structured byte-addressable NVRAM store for per-remap p2l metadata so l2p/p2l mappings remain consistent even… | ||
| CheckFreq: Frequent, Fine-Grained DNN Checkpointing | Jayashree Mohan, Amar Phanishayee, Vijay Chidambaram | 12m |
CheckFreq implements two-phase checkpointing (snapshot in GPU or CPU then asynchronous persist) plus a resumable iterator and online profiling with adaptive… | ||
| Work-in-Progress Reports (WiPs) | Keiichi Matsuzawa, Takahiro Shinagawa, Azat Nurgaliev, Marcus Paradies, Lalitha Donga, Ben Reed, Kayla Walton, Joontaek Oh, Hojin Nam, Kyoungho Koo, Youjip Won, Qiliang Li, Yinlong Xu, Min Lyu, Wei Wang, Zsolt István | 31m |
Seven short systems results: an application‑local user‑space library filesystem for persistent memory shares VM pages, omits ACLs, and in single‑process… | ||
| Facebook's Tectonic Filesystem: Efficiency from Exascale | Satadru Pan, Theano Stavrinos, Yunqiao Zhang, Atul Sikaria, Pavel Zakharov, Abhinav Sharma, Shiva Shankar P, Mike Shuey, Richard Wareing, Monika Gangapuram, Guanglei Cao, Christian Preseau, Pratap Singh, Kestutis Patiejunas, JR Tipton, Ethan Katz-Bassett, Wyatt Lloyd | 12m |
Claims a single exabyte-scale single cluster replaces many special-purpose systems, storing trillions of objects and consolidating an entire data center… | ||
| Exploiting Combined Locality for Wide-Stripe Erasure Coding in Distributed Storage | Yuchong Hu, Liangfeng Cheng, Qiaori Yao, Patrick P. C. Lee, Weichun Wang, Wei Chen | 11m |
Proposes combined locality that mixes parity locality and topology locality, constructs stripes by placing HRC-coded chunks across racks with c=f, and… | ||
| On the Feasibility of Parser-based Log Compression in Large-Scale Cloud Systems | Junyu Wei, Guangyan Zhang, Yang Wang, Zhiwei Liu, Zhanyang Zhu, Junchao Chen, Tingtao Sun, Qi Zhou | 12m |
LogReducer is a C++ reimplementation of parser-based log compression that targets numerical fields and fixes logzip bottlenecks, using delta timestamps… | ||
| CNSBench: A Cloud Native Storage Benchmark | alex marenstein | ? |
CNSBench is a Kubernetes-native benchmark that separately drives IO workloads (using existing tools) and control workloads (volume create/attach/snapshot) via… | ||
| Concordia: Distributed Shared Memory with In-Network Cache Coherence | Qing Wang, Youyou Lu, Erci Xu, Junru Li, Youmin Chen, Jiwu Shu | 12m |
Concordia implements a DSM that pushes cache coherence into programmable switches, using a hybrid protocol FLOWCC that stores per-block metadata and a… | ||
| eMRC: Efficient Miss Ratio Approximation for Multi-Tier Caching | Zhang Liu, Hee Won Lee, Yu Xiang, Dirk Grunwald, Sangtae Ha | 11m |
eMRC approximates multi-dimensional miss-ratio curves by partitioning cache space, applying a performance-cliff removal method and a convex-hull approximation… | ||
| The Storage Hierarchy is Not a Hierarchy: Optimizing Caching on Modern Storage Devices... | Kan Wu, Zhihan Guo, Guanzhou Hu, Kaiwei Tu, Ramnatthan Alagappan, Rathijit Sen, Kwanghyun Park, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau | 12m |
Proposes non‑hierarchical caching (NHC) implemented as Orthus‑CAS (block‑layer kernel module) and Orthus‑KV (user‑level for a key‑value store) that add a… | ||
| A Community Cache with Complete Information | Mania Abdi, Amin Mosayyebzadeh, Mohammad Hossein Hajkazemi, Emine Ugur Kaynar, Ata Turk, Larry Rudolph, Orran Krieger, Peter Desnoyers | 15m |
Kariz is a shared community cache that sits next to the datalake, extracts DAGs and runtime signals from Pig/Hadoop and Spark, and centrally schedules… | ||
| Learning Cache Replacement with CACHEUS | Liana V. Rodriguez, Farzana Yusuf, Steven Lyons, Eysler Paz, Raju Rangaswami, Jason Liu, Ming Zhao, Giri Narasimhan | 11m |
Classifies workloads into four primitives—LFU-friendly, LRU-friendly, scan, churn—and builds CACHEUS, an adaptive ensemble that runs experts ARC, LIRS, LFU… | ||
| FusionRAID: Achieving Consistent Low Latency for Commodity SSD Arrays | Tianyang Jiang, Guangyan Zhang, Zican Huang, Xiaosong Ma, Junyu Wei, Zhiyue Li, Weimin Zheng | 12m |
FusionRAID is a RAID redesign that implements two-phase writes (temporary replicated writes then in-place parity conversion), a dynamic block map with static… | ||
| Behemoth: A Flash-centric Training Accelerator for Extreme-scale DNNs | Shine Kim, Yunho Jin, Gina Sohn, Jonghyun Bae, Tae Jun Ham, Jae W. Lee | 12m |
Behemoth pairs a custom flash-centric memory system with matrix-accelerator nodes to hold an extreme-scale DNN on one node and enable data-parallel training… | ||
| FlashNeuron: SSD-Enabled Large-Batch Training of Very Deep Neural Networks | Jonghyun Bae, Jongsung Lee, Yunho Jin, Sam Son, Shine Kim, Hakbeom Jang, Tae Jun Ham, Jae W. Lee | 12m |
System moves intermediate activations from GPU HBM to NVMe SSDs and schedules transfers so they finish within the forward pass, using an offloading scheduler… | ||
| D2FQ: Device-Direct Fair Queueing for NVMe SSDs | Jiwon Woo, Minwoo Ahn, Gyun Lee, Jinkyu Jeong | 13m |
D2FQ offloads fair queueing to NVMe device-side WRR by mapping WRR’s three priority classes to fast/moderate/slow queues and dispatching each I/O immediately… | ||
| An In-Depth Study of Correlated Failures in Production SSD-Based Data Centers | Shujie Han, Patrick P. C. Lee, Fan Xu, Yi Liu, Cheng He, Jiongzhou Liu | 12m |
Analysis of nearly one million SSDs from Alibaba across 11 drive models shows intra-node and intra-rack correlated failures are common and group sizes… | ||
| Keynote Address: DNA Data Storage and Near-Molecule Processing for the Yottabyte Era | Luis Ceze, Karin Strauss | 01h 01m |
An end-to-end system encodes bits into ~150–300 base synthetic DNA, synthesizes pools, encapsulates them, and reads with Illumina or nanopore sequencers; the… | ||
| nothing matches | ||