--- title: EuroSys 2024 Digest venue: EuroSys year: 2024 date: '2024-01-01' tags: [] paper_count: 13 draft: false --- 13 papers selected. --- ### Pronghorn: Effective Checkpoint Orchestration for Serverless Hot-Starts *Sumer Kohli, Shreyas Kharbanda, Rodrigo Bruno, João Carreira *et al.** **TL;DR** — Demonstrates how carefully orchestrated checkpointing can eliminate cold-start latency in serverless runtimes, achieving near-instant hot-starts with negligible overhead. --- ### Serialization/Deserialization-free State Transfer in Serverless Workflows *Fangming Lu, Xingda Wei, Zhuobin Huang, Rong Chen 0001 *et al.** **TL;DR** — Eliminates the dominant serialization cost in serverless function chaining by enabling direct in-memory state passing, yielding large end-to-end latency reductions. --- ### SplitFT: Fault Tolerance for Disaggregated Datacenters via Remote Memory Logging *Xuhao Luo, Ramnatthan Alagappan, Aishwarya Ganesan* **TL;DR** — Proposes a principled fault-tolerance design for disaggregated datacenters that exploits the new memory-compute split to recover from failures with low overhead. --- ### Puddles: Application-Independent Recovery and Location-Independent Data for Persistent Memory *Suyash Mahar, Mingyao Shen, TJ Smith, Joseph Izraelevitz *et al.** **TL;DR** — Provides transparent crash consistency and data relocation for persistent memory without requiring application changes, easing adoption of PM-backed storage. --- ### Enoki: High Velocity Linux Kernel Scheduler Development *Samantha Miller, Anirudh Kumar, Tanay Vakharia, Ang Chen 0001 *et al.** **TL;DR** — Enables safe, rapid iteration on Linux CPU schedulers by isolating scheduling logic in user space while keeping kernel integration, dramatically lowering the development barrier. --- ### Transparent Multicore Scaling of Single-Threaded Network Functions *Lei Yan 0003, Yueyang Pan, Diyu Zhou, George Candea *et al.** **TL;DR** — Automatically parallelizes unmodified single-threaded network functions across cores with correctness guarantees, delivering near-linear throughput scaling without code changes. --- ### Hoda: a High-performance Open vSwitch Dataplane with Multiple Specialized Data Paths *Heng Pan, Peng He 0003, Zhenyu Li 0001, Pan Zhang *et al.** **TL;DR** — Redesigns the Open vSwitch dataplane with multiple specialized fast paths, achieving significant throughput improvements for cloud virtual networking. --- ### SmartNIC Security Isolation in the Cloud with S-NIC *Yang Zhou 0008, Mark Wilkening, James Mickens, Minlan Yu* **TL;DR** — Introduces hardware-enforced security isolation for SmartNIC offload tasks in multi-tenant clouds, preventing cross-tenant attacks without sacrificing offload performance. --- ### Finding Correctness Bugs in eBPF Verifier with Structured and Sanitized Program *Hao Sun 0021, Yiru Xu, Jianzhong Liu, Yuheng Shen *et al.** **TL;DR** — Systematically uncovers verifier logic bugs that can silently allow unsafe eBPF programs to execute in the kernel, with a structured fuzzing approach validated on the Linux eBPF verifier. --- ### CSAL: the Next-Gen Local Disks for the Cloud *Yanbo Zhou, Erci Xu, Li Zhang, Kapil Karkra *et al.** **TL;DR** — Describes a production cloud local-disk system that replaces raw NVMe with a software-defined layer to deliver better performance, reliability, and operational flexibility at hyperscale. --- ### Volley: Accelerating Write-Read Orders in Disaggregated Storage *Shaoxun Zeng, Xiaojian Liao, Hao Guo, Youyou Lu* **TL;DR** — Addresses write-read ordering hazards in disaggregated storage by pipelining acknowledgements with read-side tracking, substantially reducing tail latency without weakening durability. --- ### Erlang: Application-Aware Autoscaling for Cloud Microservices *Vighnesh Sachidananda, Anirudh Sivaraman* **TL;DR** — Leverages application-level call-graph semantics to make autoscaling decisions that are both faster and more accurate than black-box reactive policies, cutting SLO violations in microservice deployments. --- ### Automatic Root Cause Analysis via Large Language Models for Cloud Incidents *Yinfang Chen, Huaibing Xie, Minghua Ma, Yu Kang 0006 *et al.** **TL;DR** — Demonstrates an LLM-driven pipeline for automated cloud-incident root cause analysis deployed at Microsoft Azure, showing strong recall and significant reduction in on-call engineer effort.