OFC: An Original Congestion-Based Fine-grained Priority Flow Control

With the proliferation of online data intensive applications and virtualized services, the growing complexity of traffic patterns in data centers increases the likelihood of congestion, especially in incast scenarios and with a combination of short and large flows. To ensure lossless transmission, RDMA over Converged Ethernet networks rely on Priority-based Flow Control (PFC) to prevent packet loss due to buffer overflow. However, it is widely acknowledged that PFC gives rise to several issues, such as Congestion Spreading, Head-of-Line Blocking, and Deadlock, which are increasingly prominent in modern highly congested data centers. In this paper, we analyze the primary causes of congestion spreading and head-of-line blocking issues associated with PFC and propose Original congestion-based fine-grained priority Flow Control (OFC) as a solution. The performance of OFC is assessed using the programmable switch Tofino and simulations carried out with a packet-level simulator across various scenarios, encompassing incast, realistic, deadlock, and in-depth scenarios. The validation of the simulation results through testbed evaluation confirmed that OFC effectively reduces flow completion time, buffer occupancy, and deadlock occurrence by up to 60.28%, 51.47%, and 48.7%, respectively.

Authors

Institutions

Publication Details

Published
2024-12-02
DOI
https://doi.org/10.1109/secon64284.2024.10934869
Primary Topic
Scheduling and Optimization Algorithms
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

OFC: An Original Congestion-Based Fine-grained Priority Flow Control

Zhuoran Liu, Peirui Cao, Shizhen Zhao, Wenli Xiao et al.
Scheduling and Optimization Algorithms
article

OFC: An Original Congestion-Based Fine-grained Priority Flow Control

Zhuoran Liu, Peirui Cao, Shizhen Zhao, Wenli Xiao, Yuqing Yang, Xinbing Wang
article en

Abstract

With the proliferation of online data intensive applications and virtualized services, the growing complexity of traffic patterns in data centers increases the likelihood of congestion, especially in incast scenarios and with a combination of short and large flows. To ensure lossless transmission, RDMA over Converged Ethernet networks rely on Priority-based Flow Control (PFC) to prevent packet loss due to buffer overflow. However, it is widely acknowledged that PFC gives rise to several issues, such as Congestion Spreading, Head-of-Line Blocking, and Deadlock, which are increasingly prominent in modern highly congested data centers. In this paper, we analyze the primary causes of congestion spreading and head-of-line blocking issues associated with PFC and propose Original congestion-based fine-grained priority Flow Control (OFC) as a solution. The performance of OFC is assessed using the programmable switch Tofino and simulations carried out with a packet-level simulator across various scenarios, encompassing incast, realistic, deadlock, and in-depth scenarios. The validation of the simulation results through testbed evaluation confirmed that OFC effectively reduces flow completion time, buffer occupancy, and deadlock occurrence by up to 60.28%, 51.47%, and 48.7%, respectively.

Shanghai Jiao Tong University (CN)
Research Promotion Foundation
Climate action
Openalex Percentile: Top 58%
Scheduling and Optimization Algorithms
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.