Vincent Lannurien
d822cdaa6a
6.2 KiB
title, venue, year, date, tags, paper_count, draft
| title | venue | year | date | tags | paper_count | draft |
|---|---|---|---|---|---|---|
| TCC 2025 Digest | TCC | 2025 | 2025-01-01 | 12 | false |
12 papers selected.
DRKC: Deep Reinforcement Learning Enhanced Microservice Scheduling on Kubernetes Clusters in Cloud-Edge Environment
Jian Jiang, Qianmu Li, Pengchuan Wang, Yunhuai Liu
TL;DR — DRKC uses deep reinforcement learning to schedule microservices across Kubernetes clusters spanning cloud and edge nodes, optimizing latency and resource utilization.
Why notable — One of the few papers to tackle DRL-based microservice placement at the Kubernetes level in a real cloud-edge topology, making it directly actionable for practitioners.
DesFaaS: Cross-Layer Joint Dynamic Deployment System for Serverless Stateful Functions
Yuquan Jing, Binbin Feng, Zhijun Ding
TL;DR — DesFaaS jointly optimizes the placement and lifecycle of stateful serverless functions across compute, network, and storage layers to reduce latency and cost.
Why notable — Addresses the hard problem of state management in FaaS by co-designing across layers, opening a new direction for stateful serverless architectures.
CARL: Cost-Optimized Online Container Placement on VMs Using Adversarial Reinforcement Learning
Prathamesh Saraf Vinayak, Saswat Subhajyoti Mallick, Lakshmi Jagarlamudi, Anirban Chakraborty 0001 et al.
TL;DR — CARL applies adversarial reinforcement learning to online container bin-packing on cloud VMs, minimizing cost while handling adversarial workload patterns.
Why notable — The adversarial training objective makes the scheduler robust to worst-case workload shifts, a significant advance over standard RL-based placement.
CADER: Cost-Efficient Cloud Application Deployment With Tenant Requirement Guarantee in Multi-Clouds
Huaqing Tu, Ziqiang Hua, Qianpiao Ma, Hanguang Luo et al.
TL;DR — CADER places cloud application components across multiple providers to minimize cost while enforcing per-tenant SLA and data-locality constraints.
Why notable — Provides a rigorous multi-cloud placement framework that balances cost and tenant requirements, directly addressing a key challenge in multi-cloud SaaS/PaaS deployments.
Cloud Load Balancers Need to Stay Off the Data Path
Yuchen Zhang, Shuai Jin, Zhenyu Wen, Shibo He et al.
TL;DR — Based on large-scale production experience, this paper argues and demonstrates that cloud load balancers should operate out-of-band to eliminate throughput bottlenecks at scale.
Why notable — A rare production-grounded architectural insight from a major cloud provider that challenges conventional in-path load balancer designs.
PHOENIX: Misconfiguration Detection for AWS Serverless Computing
Jinfeng Wen, Haodi Ping
TL;DR — PHOENIX automatically detects security and correctness misconfigurations in AWS Lambda deployments by analyzing IAM policies, triggers, and function configurations.
Why notable — Serverless misconfiguration is a leading cause of cloud security incidents; PHOENIX provides an automated, deployable detection tool for AWS environments.
FaaSScout: Fast and Full Lifecycle RCA for FaaS Applications Using Salient Feature Mining
Min Li 0065, Jin Huang, Pengfei Chen 0002, Chongkang Tan
TL;DR — FaaSScout performs root cause analysis across the full FaaS invocation lifecycle by mining salient features from traces and logs to localize faults quickly.
Why notable — Fills a critical operational gap for serverless: fast, automated fault diagnosis that covers cold starts, platform issues, and application errors in a unified framework.
Hybrid Serverless Platform for Smart Deployment of Service Function Chains
Sheshadri K. R, J. Lakshmi
TL;DR — A hybrid serverless platform that intelligently places NFV service function chains on serverless infrastructure, reducing provisioning overhead while meeting latency targets.
Why notable — Bridges serverless computing and NFV, demonstrating that serverless abstractions can be applied to network function deployment with competitive performance.
PiCoP: Service Mesh for Sharing Microservices in Multiple Environments Using Protocol-Independent Context Propagation
Hiroya Onoe, Daisuke Kotani, Yasuo Okabe
TL;DR — PiCoP extends service mesh capabilities to span heterogeneous protocol environments by providing protocol-independent context propagation for distributed microservice tracing and control.
Why notable — Solves a practical multi-cloud and hybrid deployment challenge where microservices communicate over different protocols, enabling unified observability and policy enforcement.
A Reference Architecture for Governance of Cloud Native Applications
William Pourmajidi, Lei Zhang 0078, John Steinbacher, Tony Erwin et al.
TL;DR — Proposes and validates a reference architecture that unifies policy enforcement, compliance, and lifecycle governance for cloud-native applications across deployment environments.
Why notable — Provides a vendor-neutral governance blueprint grounded in industry practice, filling a gap between DevOps tooling and organizational cloud compliance requirements.
Observability and Incident Response in Managed Serverless Environments Using Ontology-Based Log Monitoring
Lavi Ben-Shimol, Edita Grolman, Aviad Elyashar, Inbar Maimon et al.
TL;DR — Uses an ontology-based approach to monitor serverless function logs, enabling structured incident detection and response in managed FaaS environments.
Why notable — Brings structured knowledge representation to serverless observability, enabling richer incident correlation than rule-based or purely ML-based log monitors.
A Run-Time Framework for Ensuring Zero-Trust State of Client's Machines in Cloud Environment
Devki Nandan Jha, Graham Lenton, James Asker, David Blundell et al.
TL;DR — A runtime attestation framework continuously verifies the security posture of client machines accessing cloud resources, enforcing zero-trust policies based on live system state.
Why notable — Moves zero-trust enforcement from static policy configuration to continuous runtime verification, addressing a key gap in current cloud access control models.