22nd USENIX Conference on File and Storage Technologies (FAST '24)
2024-02-27..2024-02-29
Santa Clara, United States
as the organisers put it
22nd USENIX Conference on File and Storage Technologies (FAST '24), February 26–29, 2024, Hyatt Regency Santa Clara, Santa Clara, CA, USA. FAST brings together storage-system researchers and practitioners to explore new directions in the design, implementation, evaluation, and deployment of storage systems.
who it's for
storage-system researchers, file system developers, kernel developers, cloud storage engineers, database and key-value engineers, storage performance engineers, ssd and nvme engineers
topics
RDMA-attached memory with SSD; buffered versus direct I/O chooser; collaborative near-storage caching; application and kernel cache cooperation; DRAM cache integrated with virtual memory; online miss-ratio curve generation; NVMe char io_uring passthru; file system model checking; scalable userspace filesystem framework; capacity-variant storage systems; low-latency SSD read-paths; erasure coding tiering for LSM-tree storage
| title | author | length |
|---|---|---|
| Keynote Address: Lessons Learnt in Trying to Build New Storage Technologies | Antony Rowstron | 53m |
Microsoft Research Cambridge describes a decade-long effort to replace tape/HDD for cloud archival by writing data into glass using femtosecond lasers (Project… | ||
| TeRM: Extending RDMA-Attached Memory with SSD | Zhe Yang, Qing Wang, Xiaojian Liao, Youyou Lu, Keji Huang, Jiwu Shu | 23m |
Extends RDMA-attached memory by mapping SSD-backed virtual space (64 GB VM with 32 GB RAM in tests) and replacing NIC page-fault handling with user-space… | ||
| Combining Buffered I/O and Direct I/O in Distributed File Systems | Yingjin Qian, Marc-André Vef, Patrick Farrell, Andreas Dilger, Xi Li, Shuichi Ihara, Yinjin Fu, Wei Xue, André Brinkmann | 22m |
Implements a dynamic IO-mode chooser inside Lustre that switches between buffered and direct I/O based on IO size, lock contention, and memory pressure; adds… | ||
| OmniCache: Collaborative Caching for Near-storage Accelerators | Jian Zhang, Yujie Ren, Marie Nguyen, Changwoo Min, Sudarsun Kannan | 17m |
OmniCache combines a host near-cache and a 4GB device cache (example config: 16GB host + 4GB device) with an OM index and user library to keep data in the… | ||
| Symbiosis: The Art of Application and Kernel Cache Cooperation | Yifan Dai, Jing Liu, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau | 20m |
Symbiosis dynamically sizes application and kernel caches for LevelDB, WiredTiger and RocksDB by online simulation, claiming 1.5× average and >5× peak speedups… | ||
| Optimizing File Systems on Heterogeneous Memory by Integrating DRAM Cache with Virtual... | Yubo Liu, Yuxin Ren, Mingrui Liu, Hongbo Li, Hanjun Guo, Xie Miao, Xinwei Hu, Haibo Chen | 21m |
Revisits DRAM cache use in DRAM‑PM heterogeneous memory and integrates DRAM cache with virtual memory to avoid excessive data movement, exposing a programmable… | ||
| Kosmo: Efficient Online Miss Ratio Curve Generation for Eviction Policy Evaluation | Kia Shakiba, Sari Sultan, Michael Stumm | 19m |
Kosmo generates online miss-ratio curves for non-inclusion eviction policies by keeping a single shared object copy and per-object eviction map of past… | ||
| I/O Passthru: Upstreaming a flexible and efficient I/O Path in Linux | Kanchan Joshi, Anuj Gupta, Javier González, Ankit Kumar, Krishna Kanth Reddy, Arun George, Simon Lund, Jens Axboe | 19m |
Upstream Linux receives an NVMe char interface plus an io_uring command path that adds big SQE/big CQE to avoid ioctl copies, enable non‑root access, and… | ||
| Metis: File System Model Checking via Versatile Input and State Exploration | Yifei Liu, Manish Adkar, Gerard Holzmann, Geoff Kuenning, Pei Liu, Scott A. Smolka, Wei Su, Erez Zadok | 21m |
Metis is a model-checking framework that generates system-call sequences via a nondeterministic loop and weighting scheme, stores both concrete and compact… | ||
| RFUSE: Modernizing Userspace Filesystem Framework through Scalable Kernel-Userspace... | Kyu-Jin Cho, Jaewon Choi, Hyungjoon Kwon, Jin-Soo Kim | 19m |
RFUSE implements per-core ring buffers and a hybrid busy-poll to replace FUSE’s single pending queue, preserves FUSE APIs so existing filesystems relink… | ||
| The Design and Implementation of a Capacity-Variant Storage System | Ziyang Jiao, Xiangqun Zhang, Hojin Shin, Jongmoo Choi, Bryan S. Kim | 21m |
A capacity-variant storage system (CVSS) is implemented atop Linux 5.15 with CV-FS (F2FS baseline), CV-SSD and a user-space CV-manager that coordinates… | ||
| I/O in a Flash: Evolution of ONTAP to Low-Latency SSDs | Matthew Curtis-Maury, Ram Kesavan, Bharadwaj V. R., Nikhil Mattankot, Vania Fang, Yash Trivedi, Kesari Mishra, Qin Li | 22m |
Claims that ONTAP’s read path is refactored to hit tens of microseconds with SSDs by eliminating message-passing hops and context switches. Implements MP-safe… | ||
| We Ain't Afraid of No File Fragmentation: Causes and Prevention of Its Performance... | Yuhun Jun, Shinhyun Park, Jeong-Uk Kang, Sang-Hoon Kim, Euiseong Seo | 20m |
Finds that fragmentation hurts SSD reads mainly through die-level collisions, not request splitting, because nonconsecutive die allocation from interleaved… | ||
| In-Memory Key-Value Store Live Migration with NetMigrate | Zeying Zhu, Yibo Zhao, Zaoxing Liu | 18m |
NetMigrate migrates in-memory key-value shards by pushing migration state into a top-of-rack programmable switch and rewriting client packets so queries go… | ||
| IONIA: High-Performance Replication for Modern Disk-based KV Stores | Yi Xu, Henry Zhu, Prashant Pandey, Alex Conway, Rob Johnson, Aishwarya Ganesan, Ramnatthan Alagappan | 24m |
IONIA is a storage-aware replication protocol for SSD-based write-optimized KV stores that delivers one round-trip writes by making updates durable immediately… | ||
| Physical vs. Logical Indexing with IDEA: Inverted Deduplication-Aware Index | Asaf Levi, Philip Shilane, Sarai Sheinvald, Gala Yadgar | 26m |
IDEA indexes the physical unique chunks instead of logical files, maps terms to unique chunks and maintains a chunk-to-file map, and modifies chunking to… | ||
| MIDAS: Minimizing Write Amplification in Log-Structured Systems through Adaptive Group... | Seonggyun Oh, Jeeyun Kim, Soyoung Han, Jaeho Kim, Sungjin Lee, Sam H. Noh | 21m |
MiDAS minimizes garbage-collection write amplification by organizing data into a chain of adaptive groups, automatically tuning number of groups and group… | ||
| Panel: Storage Systems in the LLM Era | Keith A. Smith, Greg Ganger, Dean Hildebrand, Glenn Lockwood, Nisha Talagala, Zhe Zhang | 50m |
Panel argues that LLM training often tolerates “bigger and cheaper” storage but surfaces concrete problems: data provenance and tracking which data or model… | ||
| What's the Story in EBS Glory: Evolutions and Lessons in Building Cloud Block Store | Weidong Zhang, Erci Xu, Qiuping Wang, Xiaolu Zhang, Yuesheng Gu, Zhenwei Lu, Tao Ouyang, Guanqun Dai, Wenwen Peng, Zhe Xu, Shuo Zhang, Dong Wu, Yilei Peng, Tianyun Wang, Haoran Zhang, Jiasheng Wang, Wenyuan Yan, Yuanyuan Dong, Wenhui Yao, Zhongjie Wu, Lingjun Zhu, Chao Shi, Yinhu Wang, Rong Liu, Junping Wu, Jiaji Zhu, Jiesheng Wu | 18m |
Three generations of Alibaba Cloud Block Store evolve from a simple two-layer replica model to an SSD/RSBD plus Pangu backend and finally to a… | ||
| ELECT: Enabling Erasure Coding Tiering for LSM-tree-based Storage | Yanjing Ren, Yuanming Ren, Xiaolu Li, Yuchong Hu, Jingwei Li, Patrick P. C. Lee | 22m |
ELECT implements hybrid redundancy on top of Cassandra, encoding LSM last-level SSTables offline into erasure-coded coding groups while keeping hot keys… | ||
| MinFlow: High-performance and Cost-efficient Data Passing for I/O-intensive Stateful... | Tao Li, Yongkun Li, Wenzhe Zhu, Yinlong Xu, John C. S. Lui | 16m |
MinFlow is a prototype framework that constructs progressively converging multi-level shuffle topologies, partitions the DAG into interleaved bipartite… | ||
| COLE: A Column-based Learned Storage for Blockchain Systems | Ce Zhang, Cheng Xu, Haibo Hu, Jianliang Xu | 21m |
COLE implements a column-based learned index that stores each state's historical values contiguously and replaces Merkle Patricia Trie nodes with… | ||
| Baleen: ML Admission & Prefetching for Flash Caches | Daniel Lin-Kit Wong, Hao Wu, Carson Molder, Sathya Gunasekar, Jimmy Lu, Snehal Khandkar, Abhinav Sharma, Daniel S. Berger, Nathan Beckmann, Gregory R. Ganger | 21m |
Baleen uses coordinated ML admission and prefetching, training on an episodes model that groups accesses from admission to eviction and optimizes an end-to-end… | ||
| Seraph: Towards Scalable and Efficient Fully-external Graph Computation via On-demand... | Tsun-Yu Yang, Yizou Chen, Yuhong Liang, Ming-Chang Yang | 21m |
Seraph is a fully-external graph engine that replaces streaming-within-grid I/O with on-demand processing tuned for modern SSDs and claims up to two orders of… | ||
| nothing matches | ||