arXiv subjects

Zili Meng

Publications and source records attributed to Zili Meng.

12 recordsLinked to original sources

arXiv2026

WiCi: Wireless GPU Computing Infrastructure

LLM inference applications are gaining significant traction. The demand for inference is growing exponentially, and the GPU usage of inference is increasingly surpassing that of training. Due to the mobility penalty, edge-side inference fails to deliver satisfactory performance. Consequently, most inference service providers currently rely on cloud-based inference, which incurs substantial, not sustainable costs for enterprises, and is even increasing in the agentic paradigm. Therefore, our goal is to enable powerful computing capabilities as server-grade GPUs on mobile devices. We propose Wireless GPU Computing Infrastructure (WiCi) in this paper. Through WiCi, mobile devices can wirelessly access server-grade GPUs, running inference tasks on mobile clients but offloading GPU-related computations to a nearby GPU via WiFi. WiCi introduces a series of designs to make sure the infrastructure is scalable with different applications, compatible with different mobile devices, and has comparable performance to running on a physical GPU. We test WiCi from mobile devices and find that WiCi can reduce time to first token by up to 90%, improve the token rate by approximately 39x compared to local inference on mobile devices for the same model, and support much larger models. WiCi also achieves up to nearly 80% of the native performance of the server-grade GPU across different applications.

cs.NI↗

arXiv2026

MD2G-Cast: Relay-Coordinated Multicast for Scalable Volumetric Streaming over MoQ

Volumetric streaming remains difficult to scale because receivers with overlapping fields of view are often served independently, causing repeated transmission of shared content. We present MD2G-Cast, a relay-coordinated multicast framework over Media over QUIC with an application-aware control layer for scalable multi-user volumetric delivery. MD2G-Cast jointly uses viewing overlap, receiver capability, and bandwidth conditions to form reusable multicast groups, share common Base content, and selectively admit Enhanced delivery. We formulate grouping and Enhanced admission as a sequential control problem, realize it with Proximal Policy Optimization (PPO), and train a compact relay model with teacher guidance for Enhanced admission. We implement MD2G-Cast with real MoQ processes and evaluate it with real access and 6DoF viewing traces for up to 100 users. At 20 and 100 users, MD2G-Cast keeps the receiver-side $P_{99}$ delivery interval below 40 ms across all seven access profiles, while Rolling reaches the 500 ms reporting cap in most cases. Across the evaluated user scales, MD2G-Cast achieves the highest or tied-highest mean system utility under homogeneous access and the highest mean utility under heterogeneous access, while reducing aggregate link load by about 27% relative to Clustering at 100 users. A matched relay-control ablation separates the control structure from its optimizer, showing that random feasible actions reduce utility while deterministic control remains competitive with PPO. Together, the results support relay coordination and selective Enhanced admission, rather than a particular policy optimizer, as the central design contribution.

eess.IV↗

arXiv2026

OCC: Physical-Layer Assisted Congestion Control for Real-Time Communications

Real-time communications (RTC) is a core technology for emerging applications in 6G, such as cloud gaming, teleoperation, and extended reality (XR), which require consistently low latency and high bitrates. Existing RTC solutions fundamentally struggle to maintain low latency while supporting high bitrates due to their reliance on trial-and-error-based mechanisms. These mechanisms fail to probe the available bandwidth (ABW) promptly and accurately, leading to a trade-off between latency reliability and bandwidth utilization. The tension becomes extremely more critical as the cellular bandwidth and application's demand fluctuate with a larger range in cellular networks nowadays. To address this trade-off, we propose OCC, a novel approach that utilizes physical-layer information to explicitly obtain the ABW in real time, enabling rapid adaptation to dynamic wireless network conditions. However, the unique characteristics of RTC, including traffic bursts, application (APP) limits, and encoder lag, make the physical-layer informed control non-trivial. OCC effectively addresses these issues through three innovative strategies: frame-aware bandwidth measurement, APP-limit-aware bandwidth estimation, and encoder-friendly rate control. Extensive over-the-air experiments on an open-source cellular testbed demonstrate that OCC significantly enhances the performance of mobile RTC, reducing tail network latency by $13\%$ to $68\%$ and improving video frame bitrate by $1.2\times$ to $3.5\times$.

cs.NI↗

arXiv2026

COMETS: Coordinated Multi-Destination Video Transmission with In-Network Rate Adaptation

Large-scale video streaming events attract millions of simultaneous viewers, stressing existing delivery infrastructures. Client-driven adaptation reacts slowly to shared congestion, while server-based coordination introduces scalability bottlenecks and single points of failure. We present COMETS, a coordinated multi-destination video transmission framework that leverages information-centric networking principles such as request aggregation and in-network state awareness to enable scalable, fair, and adaptive rate control. COMETS introduces a novel range-interest protocol and distributed in-network decision process that aligns video quality across receiver groups while minimizing redundant transmissions. To achieve this, we develop a lightweight distributed optimization framework that guides per-hop quality adaptation without centralized control. Extensive emulation shows that COMETS consistently improves bandwidth utilization, fairness, and user-perceived quality of experience over DASH, MoQ, and ICN baselines, particularly under high concurrency. The results highlight COMETS as a practical, deployable approach for next-generation scalable video delivery.

cs.NI↗

arXiv2025

Make a Video Call with LLM: A Measurement Campaign over Six Mainstream Apps

In 2025, Large Language Model (LLM) services have launched a new feature -- AI video chat -- allowing users to interact with AI agents via real-time video communication (RTC), just like chatting with real people. Despite its significance, no systematic study has characterized the performance of existing AI video chat systems. To address this gap, this paper proposes a comprehensive benchmark across four dimensions: quality, latency, internal mechanisms, and system overhead. Using custom testbeds, we further evaluate six mainstream AI video chatbots with this benchmark. We also build an online platform for user study. The measurement leads to interesting findings that could be beneficial to the future optimizations. For example, the network latency of AI video chat matters not as much as human video chat. The capabilities of AI agents matters most in the user experience. Our benchmarking results also open up several research questions for future optimizations of AI video chatbots. Availability: https://callarena.net/ for the online evaluation platform and our open-sourced dataset and testbed.

cs.NI↗

arXiv2025

INDS: Incremental Named Data Streaming for Real-Time Point Cloud Video

Real-time streaming of point cloud video, characterized by massive data volumes and high sensitivity to packet loss, remains a key challenge for immersive applications under dynamic network conditions. While connection-oriented protocols such as TCP and more modern alternatives like QUIC alleviate some transport-layer inefficiencies, including head-of-line blocking, they still retain a coarse-grained, segment-based delivery model and a centralized control loop that limit fine-grained adaptation and effective caching. We introduce INDS (Incremental Named Data Streaming), an adaptive streaming framework based on Information-Centric Networking (ICN) that rethinks delivery for hierarchical, layered media. INDS leverages the Octree structure of point cloud video and expressive content naming to support progressive, partial retrieval of enhancement layers based on consumer bandwidth and decoding capability. By combining time-windows with Group-of-Frames (GoF), INDS's naming scheme supports fine-grained in-network caching and facilitates efficient multi-user data reuse. INDS can be deployed as an overlay, remaining compatible with QUIC-based transport infrastructure as well as future Media-over-QUIC (MoQ) architectures, without requiring changes to underlying IP networks. Our prototype implementation shows up to 80% lower delay, 15-50% higher throughput, and 20-30% increased cache hit rates compared to state-of-the-art DASH-style systems. Together, these results establish INDS as a scalable, cache-friendly solution for real-time point cloud streaming under variable and lossy conditions, while its compatibility with MoQ overlays further positions it as a practical, forward-compatible architecture for emerging immersive media systems.

cs.MM↗

arXiv2025

RoTO: Robust Topology Obfuscation Against Tomography Inference Attacks

Tomography inference attacks aim to reconstruct network topology by analyzing end-to-end probe delays. Existing defenses mitigate these attacks by manipulating probe delays to mislead inference, but rely on two strong assumptions: (i) probe packets can be perfectly detected and altered, and (ii) attackers use known, fixed inference algorithms. These assumptions often break in practice, leading to degraded defense performance under detection errors or adaptive adversaries. We present RoTO, a robust topology obfuscation scheme that eliminates both assumptions by modeling uncertainty in attacker-observed delays through a distributional formulation. RoTO casts the defense objective as a min-max optimization problem that maximizes expected topological distortion across this uncertainty set, without relying on perfect probe control or specific attacker models. To approximate attacker behavior, RoTO leverages graph neural networks for inference simulation and adversarial training. We also derive an upper bound on attacker success probability, and demonstrate that our approach enhances topology obfuscation performance through the optimization of this upper bound. Experimental results show that RoTO outperforms existing defense methods, achieving average improvements of 34% in structural similarity and 42.6% in link distance while maintaining strong robustness and concealment capabilities.

cs.NI↗

arXiv2023

Confucius: Achieving Consistent Low Latency with Practical Queue Management for Real-Time Communications

Real-time communication applications require consistently low latency, which is often disrupted by latency spikes caused by competing flows, especially Web traffic. We identify the root cause of disruptions in such cases as the mismatch between the abrupt bandwidth allocation adjustment of queue scheduling and gradual congestion window adjustment of congestion control. For example, when a sudden burst of new Web flows arrives, queue schedulers abruptly shift bandwidth away from the existing real-time flow(s). The real-time flow will need several RTTs to converge to the new available bandwidth, during which severe stalls occur. In this paper, we present Confucius, a practical queue management scheme designed for offering real-time traffic with consistently low latency regardless of competing flows. Confucius slows down bandwidth adjustment to match the reaction of congestion control, such that the end host can reduce the sending rate without incurring latency spikes. Importantly, Confucius does not require the collaboration of end-hosts (e.g., labels on packets), nor manual parameter tuning to achieve good performance. Extensive experiments show that Confucius outperforms existing practical queueing schemes by reducing the stall duration by more than 50%, while the competing flows also fairly enjoy on-par performance.

cs.NI↗

arXiv2020

Interpreting Deep Learning-Based Networking Systems

While many deep learning (DL)-based networking systems have demonstrated superior performance, the underlying Deep Neural Networks (DNNs) remain blackboxes and stay uninterpretable for network operators. The lack of interpretability makes DL-based networking systems prohibitive to deploy in practice. In this paper, we propose Metis, a framework that provides interpretability for two general categories of networking problems spanning local and global control. Accordingly, Metis introduces two different interpretation methods based on decision tree and hypergraph, where it converts DNN policies to interpretable rule-based controllers and highlight critical components based on analysis over hypergraph. We evaluate Metis over several state-of-the-art DL-based networking systems and show that Metis provides human-readable interpretations while preserving nearly no degradation in performance. We further present four concrete use cases of Metis, showcasing how Metis helps network operators to design, debug, deploy, and ad-hoc adjust DL-based networking systems.

cs.NI↗

arXiv2019

Learning Scheduling Algorithms for Data Processing Clusters

Efficiently scheduling data processing jobs on distributed compute clusters requires complex algorithms. Current systems, however, use simple generalized heuristics and ignore workload characteristics, since developing and tuning a scheduling policy for each workload is infeasible. In this paper, we show that modern machine learning techniques can generate highly-efficient policies automatically. Decima uses reinforcement learning (RL) and neural networks to learn workload-specific scheduling algorithms without any human instruction beyond a high-level objective such as minimizing average job completion time. Off-the-shelf RL techniques, however, cannot handle the complexity and scale of the scheduling problem. To build Decima, we had to develop new representations for jobs' dependency graphs, design scalable RL models, and invent RL training methods for dealing with continuous stochastic job arrivals. Our prototype integration with Spark on a 25-node cluster shows that Decima improves the average job completion time over hand-tuned scheduling heuristics by at least 21%, achieving up to 2x improvement during periods of high cluster load.

cs.LG↗

arXiv2018

CoCo: Compact and Optimized Consolidation of Modularized Service Function Chains in NFV

The modularization of Service Function Chains (SFCs) in Network Function Virtualization (NFV) could introduce significant performance overhead and resource efficiency degradation due to introducing frequent packet transfer and consuming much more hardware resources. In response, we exploit the lightweight and individually scalable features of elements in Modularized SFCs (MSFCs) and propose CoCo, a compact and optimized consolidation framework for MSFC in NFV. CoCo addresses the above problems in two ways. First, CoCo Optimized Placer pays attention to the problem of which elements to consolidate and provides a performance-aware placement algorithm to place MSFCs compactly and optimize the global packet transfer cost. Second, CoCo Individual Scaler innovatively introduces a push-aside scaling up strategy to avoid degrading performance and taking up new CPU cores. To support MSFC consolidation, CoCo also provides an automatic runtime scheduler to ensure fairness when elements are consolidated on CPU core. Our evaluation results show that CoCo achieves significant performance improvement and efficient resource utilization.

cs.NI↗

arXiv2018

PAM: When Overloaded, Push Your Neighbor Aside!

Recently SmartNICs are widely used to accelerate service chains in NFV. However, when the SmartNIC is overloaded, casually migrating vNFs away from SmartNIC to CPU may lead to additional packet transmissions between SmartNIC and CPU. To address this problem, we present PAM, push aside migration to effectively alleviate the hot spot on SmartNIC with no performance overhead. Our key novelty is to push vNFs on the border of SmartNIC and CPU aside to release resources for the bottleneck vNF. Evaluation shows that PAM could efficiently alleviate the hot spot on SmartNIC and generate a service chain with much lower latency compared with the naive solution.

cs.NI↗