15th USENIX Symposium on Operating Systems Design and Implementation (OSDI '21)
2021-07-14..2021-07-16
-- online --
| title | author | length |
|---|---|---|
| Pollux: Co-adaptive Cluster Scheduling for Goodput-Optimized Deep Learning | Aurick Qiao, Sang Keun Choe, Suhas Jayaram Subramanya, Willie Neiswanger, Qirong Ho, Hao Zhang, Gregory R. Ganger, Eric P. Xing | 14m |
Pollux builds a co-adaptive scheduler that jointly models system throughput and statistical efficiency, introduces the goodput metric (their product), and… | ||
| Oort: Efficient Federated Learning via Guided Participant Selection | Fan Lai, Xiangfeng Zhu, Harsha V. Madhyastha, Mosharaf Chowdhury | 16m |
Oort is a participant-selection framework that selects clients by a combined loss-over-duration utility score, using exploration/exploitation and uncertainty… | ||
| PET: Optimizing Tensor Programs with Partially Equivalent Transformations and Automated | Haojie Wang, Jidong Zhai, Mingyu Gao, Zixuan Ma, Shizhi Tang, Liyan Zheng, Yuanzhi Li, Kaiyuan Rong, Yuanyong Chen, Zhihao Jia | 15m |
PET applies partially equivalent transformations plus automated corrections to expand the tensor-program search space and finds faster kernels by combining… | ||
| Privacy Budget Scheduling | Tao Luo, Mingen Pan, Pierre Tholoniat, Asaf Cidon, Roxana Geambasu, Mathias Lécuyer | 14m |
Built an extension to Kubernetes called PrivateKube that treats differential privacy as a non-replenishable resource and implements a scheduler DPF (Dominant… | ||
| Modernizing File System through In-Storage Indexing | Jinhyung Koo, Junsu Im, Jooyoung Song, Juhyung Park, Eunji Lee, Bryan S. Kim, Sungjin Lee | 14m |
KEVIN replaces the block interface with a host-side lightweight POSIX file system that maps files to inode-based keys (meta keys prefixed m, data keys prefixed… | ||
| Nap: A Black-Box Approach to NUMA-Aware Persistent Memory Indexes | Qing Wang, Youyou Lu, Junru Li, Jiwu Shu | 14m |
Nap converts existing concurrent persistent‑memory indexes into NUMA‑aware ones by adding a NUMA‑aware layer that keeps per-node partial views in PM for local… | ||
| Rearchitecting Linux Storage Stack for µs Latency and High Throughput | Jaehyun Hwang, Midhul Vuppalapati, Simon Peter, Rachit Agarwal | 15m |
blk-switch is a Linux-kernel rearchitecture of the block layer that maps block-layer queues to driver queues and adds per-class egress queues, prioritization… | ||
| Optimizing Storage Performance with Calibrated Interrupts | Amy Tai, Igor Smolyar, Michael Wei, Dan Tsafrir | 15m |
Optimizing Storage Performance with Calibrated Interrupts
Amy Tai, VMware Research; Igor Smolyar, Technion — Israel Institute of Technology; Michael Wei… | ||
| ZNS+: Advanced Zoned Namespace Interface for Supporting In-Storage Zone Compaction | Kyuhwa Han, Hyunho Gwak, Dongkun Shin, Jooyoung Hwang | 14m |
ZNS+ adds internal zone compaction and permits sparse sequential writes so the host offloads data-copy for segment reclamation to the SSD, implements… | ||
| DMon: Efficient Detection and Correction of Data Locality Problems Using Selective | Tanvir Ahmed Khan, Ian Neal | 14m |
Selective profiling collects runtime data selectively and incrementally to locate and repair data-locality bugs with low overhead; the prototype DMon… | ||
| CLP: Efficient and Scalable Search on Compressed Text Logs | Kirk Rodrigues, Yu Luo, Ding Yuan | 15m |
A system called CLP losslessly compresses text logs and stores 14 TB as 328 GB (2.26% of original) and reports an average compression ratio of 32.2 versus gzip… | ||
| Polyjuice: High-Performance Transactions via Learned Concurrency Control | Jiachen Wang, Ding Ding, Huan Wang, Conrad Christensen, Zhaoguo Wang, Haibo Chen | 15m |
Polyjuice is a learned concurrency-control system that encodes CC as a per-access policy table mapping (transaction type, access id) to choices for waiting… | ||
| Retrofitting High Availability Mechanism to Tame Hybrid Transaction/Analytical Processing | Sijie Shen, Rong Chen, Haibo Chen, Binyu Zang | 14m |
Vegito retrofits replica-based HA so backups serve analytical queries while primaries handle OLTP, using a gossip-style epoch oracle to give consistent… | ||
| The nanoPU: A Nanosecond Network Stack for Datacenters | Stephen Ibanez, Alex Mallery, Serhat Arslan, Theo Jepsen, Muhammad Shahbaz, Changhoon Kim, Nick McKeown | 15m |
Prototype integrates NIC and CPU on FPGA, adds a register-file fast path that places incoming messages into CPU registers, and a hardware thread scheduler that… | ||
| Beyond malloc efficiency to fleet efficiency: a hugepage-aware memory allocator | A.H. Hunter, Chris Kennelly, Paul Turner, Darryl Gove, Tipp Moseley, Parthasarathy Ranganathan | 14m |
An enhancement of TCMALLOC called TEMERAIRE implements a huge-filler bin-packer that allocates in hugepage-sized spans, tracks per-hugepage metadata (longest… | ||
| Scalable Memory Protection in the PENGLAI Enclave | Erhu Feng, Xu Lu, Dong Du, Bicheng Yang, Xueqiang Jiang, Yubin Xia, Binyu Zang, Haibo Chen | 14m |
Prototype Penglai adds two hardware primitives, Guarded Page Table (GPT) and Mountable Merkle Tree (MMT) plus a shadow fork fast-init, enabling page-grained… | ||
| NrOS: Effective Replication and Sharing in an Operating System | Ankit Bhardwaj, Chinmay Kulkarni, Reto Achermann, Irina Calciu, Sanidhya Kashyap, Ryan Stutsman, Amy Tai, Gerd Zellweger | 15m |
NrOS implements a sequential kernel per NUMA node and keeps a replica per node, synchronizing replicas with an operation log with combiner batching so cores… | ||
| Addra: Metadata-private voice communication over fully untrusted infrastructure | Ishtiyaque Ahmad, Yuntian Yang, Divyakant Agrawal, Amr El Abbadi, Trinabh Gupta | 15m |
Addra implements a mailbox architecture where senders deposit outgoing packets to a fixed mailbox and receivers retrieve them with two-hop mailbox assignment… | ||
| Bringing Decentralized Search to Decentralized Services | Mingyu Li, Jinhao Zhu, Tianxu Zhang, Cheng Tan, Yubin Xia, Sebastian Angel, Haibo Chen | 14m |
DeSearch builds a decentralized, verifiable, private search pipeline for Steemit and OpenBazaar, using trusted hardware (SGX) and ORAM for privacy, cloud… | ||
| Finding Consensus Bugs in Ethereum via Multi-transaction Differential Fuzzing | Youngseok Yang, Taesoo Kim, Byung-Gon Chun | 15m |
Fluffy mutates and executes multi-transaction differential fuzzing test cases against multiple Ethereum clients (Geth, OpenEthereum), exercising different… | ||
| MAGE: Nearly Zero-Cost Virtual Memory for Secure Computation | Sam Kumar, David E. Culler, Raluca Ada Popa | 15m |
MAGE implements a planner and runtime that precomputes the memory programming schedule for oblivious secure computation and applies Belady's algorithm plus… | ||
| Zeph: Cryptographic Enforcement of End-to-End Data Privacy | Lukas Burkhalter, Nicolas Küchler, Alexander Viand, Hossein Shafagh, Anwar Hithnawi | 15m |
Zeph enforces user-defined privacy by encrypting data at sources and issuing cryptographic transformation tokens from a separate privacy plane so services can… | ||
| DistAI: Data-Driven Automated Invariant Learning for Distributed Protocols | Jianan Yao, Runzhou Tao, Ronghui Gu, Jason Nieh, Suman Jana, Gabriel Ryan | 14m |
DistAI simulates protocols to collect states, enumerates strongest possible invariants up to a bounded formula size, and checks inductiveness with an SMT-based… | ||
| GoJournal: a verified, concurrent, crash-safe journaling system | Tej Chajed, Joseph Tassarotti, Mark Theng, Ralf Jung, M. Frans Kaashoek, Nickolai Zeldovich | 14m |
GoJournal implements a verified, concurrent, crash-safe journaling API in Go and proves it in Coq with Perennial 2.0, using the Goose translator to model Go… | ||
| STORM: Refinement Types for Secure Web Applications | Nico Lehmann, Rose Kunkel, Jordan Brown, Jean Yang, Niki Vazou, Nadia Polikarpova, Deian Stefan, Ranjit Jhala | 15m |
Storm is a web framework that enforces data-dependent security policies at compile time via a Security Typed ORM that refines MVC types with logical… | ||
| Horcrux: Automatic JavaScript Parallelism for Resource-Efficient Web Computation | Shaghayegh Mardani, Ayush Goel, Ronny Ko, Harsha V. Madhyastha, Ravi Netravali | 14m |
Horcrux automatically parallelizes legacy page JavaScript by doing offline concolic + dynamic analysis on servers to produce conservative per-function… | ||
| SANRAZOR: Reducing Redundant Sanitizer Checks in C/C++ Programs | Jiang Zhang, Shuai Wang, Manuel Rigger, Pinjia He, Zhendong Su | 14m |
SanRazor is a tool that identifies and removes redundant sanitizer checks by combining dynamic coverage profiles with static data-dependency trees, instruments… | ||
| Dorylus: Affordable, Scalable, and Accurate GNN Training with Distributed CPU Servers and | John Thorpe, Yifan Qiao, Jonathan Eyolfson, Shen Teng, Guanzhou Hu, Zhihao Jia, Jinliang Wei, Keval Vora, Ravi Netravali, Miryung Kim, Guoqing Harry Xu | 15m |
Dorylus: Affordable, Scalable, and Accurate GNN Training with Distributed CPU Servers and Serverless Threads
John Thorpe, Yifan Qiao, Jonathan Eyolfson, and… | ||
| GNNAdvisor: An Adaptive and Efficient Runtime System for GNN Acceleration on GPUs | Yuke Wang, Boyuan Feng, Gushu Li, Shuangchen Li, Lei Deng, Yuan Xie, Yufei Ding | 13m |
GNNAdvisor implements an adaptive GPU runtime that uses 2D workload management and community-aware node remapping plus a lightweight analytical model for… | ||
| Marius: Learning Massive Graph Embeddings on a Single Machine | Jason Mohoney, Roger Waleffe, Henry Xu, Theodoros Rekatsinas, Shivaram Venkataraman | 13m |
A system called Marius trains billion-edge graph embeddings on a single machine by combining partition buffer, asynchronous I/O, pipelining and a new beta… | ||
| P3: Distributed Deep Graph Learning at Scale | Swapnil Gandhi, Anand Padmanabha Iyer | 14m |
P3 (pcube) is a distributed GNN runtime that separates graph structure from features, never ships full feature vectors across the network and instead pushes… | ||
| Distributed Trust: Is “Blockchain” the answer? | Radia Perlman | 54m |
Argues that blockchain often isn’t the right tool and backs it with specifics: bitcoin’s proof-of-work targets a new block every ~10 minutes and now consumes… | ||
| It's Time for Operating Systems to Rediscover Hardware | Timothy Roscoe | 01h 06m |
Argument: mainstream OS research treats hardware as a uniform cache‑coherent SMP and ignores the heterogeneous reality, with concrete consequences: the… | ||
| AI in Finance: Scope and Examples | Manuela Veloso | 59m |
Image-based models convert time series into pictures with a Mondrian classifier (S&P500 historical data) that distinguishes buy/no-buy and an autoencoder that… | ||
| nothing matches | ||