arXiv subjects

Zuqing Zhu

Publications and source records attributed to Zuqing Zhu.

7 recordsLinked to original sources

arXiv2026

RIGEL: Real-time Optical Anomaly Diagnosis with Stateful In-Network Inference based on Distributed On-switch GNNs

The recent booming of data-intensive applications has complicated optical network management, making real-time optical anomaly diagnosis a must-have feature. However, existing approaches are mostly based on centralized data analytics and thus can hardly avoid the latency and overhead due to message exchanges between data and control planes. In this work, we propose and prototype RIGEL, which, to the best of our knowledge, is the first real-time optical anomaly diagnosis system that realizes stateful distributed in-network inference through collaborative graph neural networks (GNNs) on Tofino switches. The system is designed to be fully in-network, and a software-hardware co-design is proposed to preprocess high-dimensional spectral data for being suitable for hardware-based in-network inference. Next, we first develop an effective model to combine an autoencoder with a GraphSAGE-based GNN, and then propose a generalizable method to adapt the model to Tofino switch. The effectiveness of RIGEL is showcased in a realistic packet-over-optical network testbed, verifying that it achieves highly accurate diagnosis to detect and locate optical anomalies timely and highlighting its benefits over the state-of-the-art methods.

cs.NI↗

arXiv2026

Rain: RDMA-assisted In-Network Scheduling for Microsecond-scale Workloads

Modern data center applications increasingly require microsecond-scale service time with strict tail latency requirements, which can hardly be realized with existing in-network task schedulers due to their inherent limitations. Specifically, software-based schedulers struggle to balance throughput and latency, while switch-based designs either lack global coordination, rely on packet recirculation heavily, or only offer limited support for large tasks. In light of these restrictions of the state-of-the-arts (SOTAs), we, in this work, propose Rain, an RDMA-assisted in-network scheduler built atop programmable switches that maintains centralized queues while bounding worker-local queues. Rain introduces a bidirectional on-switch queuing mechanism to buffer and match tasks and worker-issued tokens directly in the switch, avoiding worker-side polling and approximating the optimal behavior of join-bounded-shortest-queue without global aggregation. A switch-driven RDMA engine pre-writes arbitrarily large tasks via one-sided WRITE multicasts, keeping only compact metadata on the switch. Slice-aware scheduling further localizes decisions to more homogeneous queues, reducing dispersion-induced head-of-line blocking. Moreover, our study reveals that real-world systems can diverge from theoretical predictions: shallower worker queues do not always improve tail latency. Leveraging this insight, Rain incorporates an adaptive scheduling strategy to optimize worker queue depths and worker-to-slice mappings at runtime. Evaluations with the real-world application RocksDB show that Rain achieves 1.75x higher throughput than the best-performing SOTA while satisfying the same tail latency requirement.

cs.NI↗

arXiv2021

Passive Optical Networking for 5G and Beyond 5G Low-Latency Mobile Fronthauling Services

Passive optical network (PON) technology offers an attractive cost-efficient alternative to support 5G and Beyond 5G mobile network fronthauling (MFH). However, MFH for such networks is challenging given its high bandwidth and strict latency requirements. To reduce these requirements, radio access network (RAN) functional splitting has been introduced in 5G networks; this provides more flexibility in resource allocation since the protocol stack is distributed between the centralized and the distributed units. In contrast to the conventional MFH requirement of the RF-PHY splitting, the MFH traffic produced by higher-layer splittings becomes more dependent on the actual user traffic load. By capitalizing on the new characteristics of the MFH traffic with RAN functional splitting, this article introduces a resource allocation mechanism to improve the performance of PONs serving MFH.

cs.NI↗

arXiv2021

Federated Learning over Next-Generation Ethernet Passive Optical Networks

Federated Learning (FL) is a distributed machine learning (ML) type of processing that preserves the privacy of user data, sharing only the parameters of ML models with a common server. The processing of FL requires specific latency and bandwidth demands that need to be fulfilled by the operation of the communication network. This paper introduces a Dynamic Wavelength and Bandwidth Allocation algorithm for Quality of Service (QoS) provisioning for FL traffic over 50 Gb/s Ethernet Passive Optical Networks. The proposed algorithm prioritizes FL traffic and reduces the delay of FL and delay-critical applications supported on the same infrastructure.

cs.NI↗

arXiv2020

On Throughput Optimization and Bound Analysis in Cache-Enabled Fiber-Wireless Networks

With the dense deployment of millimeter wave (mmWave) front ends and popularization of bandwidth-intensive applications, shared backhaul in fiber-wireless (FiWi) networks is still facing a bandwidth crunch. To alleviate the backhaul pressure, in this paper, caching capability is enabled at the edge of FiWi networks, i.e., optical network unit access points (ONU-APs). On the other hand, as both power budget and backhaul bandwidth in FiWi networks are constrained, it is challenging to properly leverage power for caching and that for wireless transmission to achieve superior system performance. As caching has a significant impact on resource allocation, we reconsider performance optimization and analysis in cache-enabled FiWi networks. Firstly, in the cache-enabled FiWi network with mmWave, we formulate the joint power allocation and caching problem, with the goal to maximize the downlink throughput. A two-stage algorithm is then proposed to solve the problem. Secondly, to investigate the theoretical capacity of the cache-enabled FiWi network with mmWave, we derive an upper bound of the downlink throughput by analyzing properties of the average rate of wireless links. Particularly, we show that appropriate power allocation for wireless transmission and caching at ONU-APs is essential to achieve higher throughput. The numerical and simulation results validate our theoretical analysis and demonstrate the proposed algorithm can approach the analytical upper bound.

cs.NI↗

arXiv2019

DeepRMSA: A Deep Reinforcement Learning Framework for Routing, Modulation and Spectrum Assignment in Elastic Optical Networks

This paper proposes DeepRMSA, a deep reinforcement learning framework for routing, modulation and spectrum assignment (RMSA) in elastic optical networks (EONs). DeepRMSA learns the correct online RMSA policies by parameterizing the policies with deep neural networks (DNNs) that can sense complex EON states. The DNNs are trained with experiences of dynamic lightpath provisioning. We first modify the asynchronous advantage actor-critic algorithm and present an episode-based training mechanism for DeepRMSA, namely, DeepRMSA-EP. DeepRMSA-EP divides the dynamic provisioning process into multiple episodes (each containing the servicing of a fixed number of lightpath requests) and performs training by the end of each episode. The optimization target of DeepRMSA-EP at each step of servicing a request is to maximize the cumulative reward within the rest of the episode. Thus, we obviate the need for estimating the rewards related to unknown future states. To overcome the instability issue in the training of DeepRMSA-EP due to the oscillations of cumulative rewards, we further propose a window-based flexible training mechanism, i.e., DeepRMSA-FLX. DeepRMSA-FLX attempts to smooth out the oscillations by defining the optimization scope at each step as a sliding window, and ensuring that the cumulative rewards always include rewards from a fixed number of requests. Evaluations with the two sample topologies show that DeepRMSA-FLX can effectively stabilize the training while achieving blocking probability reductions of more than 20.3% and 14.3%, when compared with the baselines.

cs.NI↗

arXiv2018

How the Network Topology, Traffic Distribution, and Routing Scheme Impact on the Spectrum Usage in Elastic Optical Networks

Elastic Optical Network (EON) has been considered as a promising optical networking technology to architect the next-generation backbone networks. Routing and Spectrum Assignment (RSA) is the fundamental problem in EONs to realize service provisioning. Generally, the RSA is solved by routing the requests with lightpaths first and then assigning spectrum resources to the lightpaths to optimize the spectrum usage. Thus, the spectrum assignment explicitly decide the final spectrum usage of EONs. However, besides the spectrum assignment, there are three other factors, the network topology, traffic distribution and routing scheme, implicitly impact on the spectrum usage. Few related work involves in the implicit impact mechanism. In this paper, we aim to provide a thoroughly theoretical analysis on the impact of the three key factors on the spectrum usage. To this end, two theoretical chains are proposed: (1) The optimal spectrum usage can be measured by the chromatic number of the conflict graph, which is positively correlated to the intersecting probability, \emph{i.e.}, the smaller the intersecting probability, the smaller the optimal spectrum usage; (2) The intersecting probability is decided by the network topology, traffic distribution and routing scheme via a quadratic programming parameterized with a matrix of conflict coefficients. The effectiveness of our theoretical analysis has been validated by extensive numerical results. Meanwhile, our theoretical deductions also permit to give several constant approximation ratios for RSA algorithms.

cs.NI↗