Vincent Lannurien
d822cdaa6a
4.2 KiB
title, venue, year, date, tags, paper_count, draft
| title | venue | year | date | tags | paper_count | draft |
|---|---|---|---|---|---|---|
| TOCS 2024 Digest | TOCS | 2024 | 2024-01-01 | 8 | false |
8 papers selected.
PMAlloc: A Holistic Approach to Improving Persistent Memory Allocation
Zheng Dang, Shuibing He, Xuechen Zhang, Peiyi Hong et al.
TL;DR — PMAlloc redesigns persistent memory allocation end-to-end, co-optimizing the allocator's data structures, concurrency, and crash consistency to dramatically reduce allocation overhead.
Why notable — Persistent memory is still poorly understood at the allocator level; this paper offers a rare holistic treatment that will inform future PM software stacks.
Boki: Towards Data Consistency and Fault Tolerance with Shared Logs in Stateful Serverless Computing
Zhipeng Jia, Emmett Witchel
TL;DR — Boki introduces a shared-log abstraction for serverless functions that provides strong consistency and fault tolerance without requiring developers to manage state explicitly.
Why notable — It reframes stateful serverless as a log-centric problem, offering a clean systems primitive that substantially simplifies correctness guarantees in function-as-a-service platforms.
Diciclo: Flexible User-level Services for Efficient Multitenant Isolation
Giorgos Kappes, Stergios V. Anastasiadis
TL;DR — Diciclo provides a user-level framework that lets services customize their isolation mechanisms without kernel modifications, reducing interference among co-located tenants.
Why notable — Multitenant isolation in cloud systems is typically a blunt instrument; this work shows that flexible, low-overhead isolation can be achieved entirely in user space.
SPATA: Effective OS Bug Detection with Summary-Based, Alias-Aware, and Path-Sensitive Typestate Analysis
Tuo Li, Jia-Ju Bai, Yulei Sui, Shi-Min Hu
TL;DR — SPATA applies a summary-based, alias-aware, and path-sensitive typestate analysis to detect resource-management bugs in OS kernels at scale.
Why notable — Finding use-after-free and double-free bugs in OS code remains an open challenge; SPATA's precision improvements over prior static analyses make it a practical tool for kernel hardening.
Trinity: High-Performance and Reliable Mobile Emulation through Graphics Projection
Hao Lin, Zhenhua Li, Di Gao, Yunhao Liu et al.
TL;DR — Trinity projects GPU rendering workloads from a mobile device onto a remote high-performance GPU, enabling faithful, high-throughput mobile emulation.
Why notable — Mobile app testing at scale requires accurate emulation of GPU behavior; Trinity's graphics-projection design closes a long-standing fidelity gap in mobile emulators.
Optimizing Resource Management for Shared Microservices: A Scalable System Design
Shutian Luo, Chenyu Lin, Kejiang Ye, Guoyao Xu et al.
TL;DR — This paper presents a scalable resource-management system for shared microservices that reduces interference and improves utilization in large-scale production deployments.
Why notable — Microservice co-location is the norm in modern clouds, yet managing their shared resources at scale remains unsolved; this work delivers practical, production-validated answers.
Hardware-Software Collaborative Tiered-Memory Management Framework for Virtualization
Sai Sha, Chuandong Li, Xiaolin Wang, Zhenlin Wang et al.
TL;DR — A hardware-software co-design framework that transparently manages hot/cold data placement across DRAM and slower memory tiers inside virtual machines.
Why notable — As CXL-attached and NVM memory tiers become mainstream in data centers, principled tiered-memory management in hypervisors becomes critical; this paper provides a solid baseline.
Component-distinguishable Co-location and Resource Reclamation for High-throughput Computing
Laiping Zhao, Yushuai Cui, Yanan Yang, Xiaobo Zhou et al.
TL;DR — This system differentiates micro-components of co-located workloads to reclaim idle resources precisely, boosting overall cluster throughput without violating SLOs.
Why notable — Coarse-grained co-location wastes significant cluster capacity; the component-level granularity introduced here sets a new standard for resource-reclamation systems.