Shizhen Zhao’s team presents the Latest Research Result at NSDI 2023

Recently, the top computer networking conference USENIX NSDI 2023 was held in Boston, USA. This year marks the 20th anniversary of the establishment of NSDI, and it is also the first international event in the field of computer networking systems to be held entirely offline since 2020. Shizhen Zhao, an associate professor at the John Hopcroft Center for Computer Science in SEIEE, was invited to attend and report on his team's latest research findings. Shizhen Zhao’s team has designed a network architecture called "Flattened Clos" that can eliminate the PFC-induced deadlocks in expander networks. The related paper "Flattened Clos: Designing High-performance Deadlock-free Expander Data Center Networks Using Graph Contraction" has been accepted by the conference. All authors of the paper are from Shanghai Jiao Tong University, and Shizhen Zhao and his master's student Qizhou Zhang are co-first authors.


The USENIX Symposium on Networked Systems Design and Implementation (NSDI) is a renowned top conference in the field of computer networking systems. It focuses on the design, implementation, and testing of networks and distributed systems, emphasizing interdisciplinary and practical approaches, and attracting significant attention from both academia and industry. The conference paper selection process is rigorous, with an acceptance rate of only 16.0% for NSDI 2023. The accepted papers represent the cutting-edge research in the field of computer networking systems internationally.


Clos networks have witnessed the successful deployment of RoCE in production data centers. However, as DCN bandwidth keeps increasing, building Clos networks is becoming cost-prohibitive and thus the more cost-efficient expander graph has received much attention in recent literature. Unfortunately, the existing expander graphs' topology and routing designs may contain Cyclic Buffer Dependency (CBD) and incur deadlocks in PFC-enabled RoCE networks.

We propose Flattened Clos (FC), a topology/routing codesigned approach, to eliminate the PFC-induced deadlocks in expander networks. FC's topology and routing are designed in three steps: 1) logically divide each ToR switch into k virtual layers and establish connections only between adjacent virtual layers; 2) generate virtual up-down paths for routing; 3) flatten the virtual multi-layered network and the virtual up-down paths using graph contraction. We rigorously prove that FC's design is deadlock-free and validate this property using a real testbed and packet-level simulation. Compared to expander graphs with the edge-disjoint-spanning-tree (EDST) based routing (a state-of-art CBD-free routing algorithm for expander graphs), FC reduces the average hop count by at least 50% and improves network throughput by 2−10× or more. Compared to Clos networks with up-down routing, FC increases network throughput by 1.1−2× under all-to-all and uniform random traffic patterns.


Paper link:

Shizhen Zhao is an associate professor at Shanghai Jiao Tong University. He earned his Ph.D. from Purdue University in 2015. From 2015 to early 2019, he worked at Google's networking group in the United States. Since 2019, he has been a member of the John Hopcroft Center for Computer Science at Shanghai Jiao Tong University. His research mainly focuses on the network architecture of optoelectronic hybrid data centers, and his related research has been published in internationally renowned conferences and journals such as NSDI, SIGMETRICS, ICNP, INFOCOM, and TON.

[ 2023-05-06 ]