Difference between revisions of "Resource:Previous Seminars"

From MobiNetS
Jump to: navigation, search
 
(2 intermediate revisions by the same user not shown)
Line 1: Line 1:
=== History ===
=== History ===
{{Hist_seminar
{{Hist_seminar
|abstract = Running deep neural networks (DNNs) on large-scale videos from widely distributed cameras presents two significant challenges. Firstly, video quality for analytical purposes is severely impacted by the camera deployment environment, which is termed Pixel Recession in this paper. Secondly, low-latency video streaming from the source camera to edge servers is greatly hindered by the rapid expansion of video traffic. Despite numerous efforts such as enhancing the video structure, uneven encoding, and filtering frames captured on camera, these methods have proven insufficient to address the challenges at hand. We propose Spliceosome, a novel video analytics system that effectively overcomes the pixel recession and streaming bottlenecks. In brief, Spliceosome 1) recovers from pixel recession by adaptive video knobs (i.e., brightness and contrast) tuning in ARP (anchor region proposal) granularity, and 2) lowers the transmission volume by video thinning, which uses only single-channel information for video encoding. We implemented Spliceosome using only commercial off-the-shelf hardware. Our experimental results demonstrate that Spliceosome outperforms other alternative designs by 4.71-14.47%, 40.94-58.71%, and 14.28% in detection accuracy, end-to-end delay, and efficiency of DNNs inference, respectively.
|confname =ToN'25
|link = https://ieeexplore.ieee.org/abstract/document/10843977
|title= Spliceosome: On-Camera Video Thinning and Tuning for Timely and Accurate Analytics
|speaker=Zhongwei Sun
|date=2025-11-28
}}{{Hist_seminar
|abstract =The rapid expansion of large language models (LLMs) requires the development of extensive GPU clusters, with companies deploying clusters with tens to hundreds of thousands of GPUs. This growth significantly expands the design space for LLM training systems, requiring thorough exploration of different parallelization strategies, communication parameters, congestion control, fabric topology, etc. Current methods require up to 10k simulation experiments to identify optimal configurations, with inadequate exploration leading to significant degradation of training performance. In this paper, we tackle the overlooked problem of efficiently conducting parallel simulation experiments for design space exploration. Our {{Hist_seminar
|abstract = As Large Language Models (LLMs) continue to scale, optimizing their deployment requires efficient hardware and system co-design. However, current LLM performance evaluation frameworks fail to capture both chip-level execution details and system-wide behavior, making it difficult to assess realistic performance bottlenecks. In this work, we introduce ReaLLM, a trace-driven simulation framework designed to bridge the gap between detailed accelerator design and large-scale inference evaluation. Unlike prior simulators, ReaLLM integrates kernel profiling derived from detailed microarchitectural simulations with a new trace-driven end-to-end system simulator, enabling precise evaluation of parallelism strategies, batching techniques, and scheduling policies. To address the high computational cost of exhaustive simulations, ReaLLM constructs a precomputed kernel library based on hypothesized scenarios, interpolating results to efficiently explore a vast design space of LLM inference systems. Our validation against real hardware demonstrates the framework's accuracy, achieving an average end-to-end latency prediction error of only 9.1% when simulating inference tasks running on 4 NVIDIA H100 GPUs. We further use ReaLLM to evaluate popular LLMs' end-to-end performance across traces from different applications and identify key system bottlenecks, showing that modern GPU-based LLM inference is increasingly compute-bound rather than memory-bandwidth bound at large scale. Additionally, we significantly reduce simulation time with our precomputed kernel library by a factor of 6× for full-simulations and 164× for workload SLO exploration. ReaLLM is open-source and available at https://github.com/bespoke-silicon-group/reallm..
|confname =ASAP'25
|link = https://ieeexplore.ieee.org/abstract/document/11113621
|title= ReaLLM: A Trace-Driven Framework for Rapid Simulation of Large-Scale LLM Inference
|speaker=JunZhe
|date=2025-11-21
}}{{Hist_seminar
|abstract =With the proliferation of mobile devices, spatial crowdsourcing has emerged as a promising paradigm for facilitating location-based services, encompassing various applications across academia and industries. Recently, pioneering works have attempted to infer workers' mobility patterns from historical data to improve the quality of task assignment. However, these studies have overlooked or under-examined issues such as the dynamic mobility patterns of crowd workers, especially in the context of newcomers, the misalignment between the objectives of mobility prediction and task assignment, and the effective utilization of predicted mobility patterns. In this paper, we investigate a problem we term Task Assignment in Mobility Prediction-aware Spatial Crowdsourcing (TAMP). To address the TAMP problem, we first propose a task-adaptive meta-learning algorithm, which trains a set of specific meta-knowledge for workers' mobility prediction models through game theory-based learning task clustering and meta-training within each cluster. Then, we design a task assignment-oriented loss function and develop a task assignment algorithm that incorporates prediction performance, prioritizing assignments with higher confidence of completion. Extensive experiments on real-world datasets validate that our proposed methods can effectively improve the quality of task assignment.
|confname =ICDE'25
|link = https://ieeexplore.ieee.org/document/11113007
|title= Effective Task Assignment in Mobility Prediction-Aware Spatial Crowdsourcing
|speaker= Zhenguo
|date=2025-11-21
}}{{Hist_seminar
|abstract = Entanglement distribution across remote distances is critical for many quantum applications. Currently, the de facto approach for remote entanglement distribution relies on optical fiber for on-the-ground entanglement distribution. However, the fiber-based approach is incapable of global-scale entanglement distribution due to intrinsic limitations. This paper investigates a new hybrid ground-satellite quantum network architecture (QuESat) for global-scale entanglement distribution, integrating an on-the-ground fiber network with a global-scale passive optical network built with low-Earth-orbit satellites. The satellite network provides dynamic construction of photon lightpaths based on near-vacuum beam guides constructed via adjustable arrays of lenses, forwarding photons from one ground station to another with very high efficiency over long distances compared to using fiber. To assess the feasibility and effectiveness of QuESat for global communication, we formulate lightpath provisioning and entanglement distribution problems, considering the orbital dynamics of satellites and the time-varying entanglement demands from ground users. A two-stage algorithm is developed to dynamically configure the beam guides and distribute entanglements, respectively. The algorithm combines randomized and deterministic rounding for lightpath provisioning to enable global connectivity, with optimal entanglement swapping for distributing entanglements to meet users' demands. By developing a ground-satellite quantum network simulator, QuESat achieves multi-fold improvements compared to repeater networks.
|confname = INFOCOM'25
|link = https://ieeexplore.ieee.org/document/11044649
|title= QuESat: Satellite-Assisted Quantum Internet for Global-Scale Entanglement Distribution
|speaker= Yaliang
|date=2025-11-07
}}{{Hist_seminar
|abstract =The global business of transnational enterprises demands geo-distributed databases, where the leader-follower-based consensus protocols are the key to guaranteeing consistency of replicas spread across regions. Compared with traditional databases running in a single data center, determining which node is the leader in consensus protocol has a greater per-formance impact in geo-distributed databases running across multiple data centers. However, the performance of legacy leader management is far from satisfactory due to the network and application dynamics (e.g., network delay, node popularity, operation read-write ratio). This paper proposes GeoLM toward performance-oriented leader management for geo-distributed consensus protocols. GeoLM captures the network and application dynamics and proactively conducts seamless leader handovers with bounded switching costs. Our geo-distributed experimental results show that GeoLM improves performance up to 49.75% over the baselines (e.g., Raft and Geo-Raft) and achieves considerably good performance compared to state-of-the-art consensus protocols (e.g., SwiftPaxos, CURP, and EPaxos).
|confname = INFOCOM'25
|link = https://ieeexplore.ieee.org/document/11044598
|title= GeoLM: Performance-oriented Leader Management for Geo-Distributed Consensus Protocol
|speaker= Linqi Liu
|date=2025-11-07
}}{{Hist_seminar
|abstract = Immersive telepresence has the potential to revolutionize remote communication by offering a highly interactive and engaging user experience. However, state-of-the-art exchanges large volumes of 3D content to achieve satisfactory visual quality, resulting in substantial Internet bandwidth consumption. To tackle this challenge, we introduce MagicStream, a first-of-its-kind semantic-driven immersive telepresence system that effectively extracts and delivers compact semantic details of captured 3D representation of users, instead of traditional bit-by-bit communication of raw content. To minimize bandwidth consumption while maintaining low end-to-end latency and high visual quality, MagicStream incorporates the following key innovations: (1) efficient extraction of user's skin/cloth color and motion semantics based on lighting characteristics and body keypoints, respectively; (2) novel, real-time human body reconstruction from motion semantics; and (3) on-the-fly neural rendering of users' immersive representation with color semantics. We implement a prototype of MagicStream and extensively evaluate its performance through both controlled experiments and user trials. Our results show that, compared to existing schemes, MagicStream can drastically reduce Internet bandwidth usage by up to 1195X while maintaining good visual quality.
|abstract = Immersive telepresence has the potential to revolutionize remote communication by offering a highly interactive and engaging user experience. However, state-of-the-art exchanges large volumes of 3D content to achieve satisfactory visual quality, resulting in substantial Internet bandwidth consumption. To tackle this challenge, we introduce MagicStream, a first-of-its-kind semantic-driven immersive telepresence system that effectively extracts and delivers compact semantic details of captured 3D representation of users, instead of traditional bit-by-bit communication of raw content. To minimize bandwidth consumption while maintaining low end-to-end latency and high visual quality, MagicStream incorporates the following key innovations: (1) efficient extraction of user's skin/cloth color and motion semantics based on lighting characteristics and body keypoints, respectively; (2) novel, real-time human body reconstruction from motion semantics; and (3) on-the-fly neural rendering of users' immersive representation with color semantics. We implement a prototype of MagicStream and extensively evaluate its performance through both controlled experiments and user trials. Our results show that, compared to existing schemes, MagicStream can drastically reduce Internet bandwidth usage by up to 1195X while maintaining good visual quality.
|confname = Sensys'24
|confname = Sensys'24

Latest revision as of 20:28, 4 December 2025

History

|abstract =The rapid expansion of large language models (LLMs) requires the development of extensive GPU clusters, with companies deploying clusters with tens to hundreds of thousands of GPUs. This growth significantly expands the design space for LLM training systems, requiring thorough exploration of different parallelization strategies, communication parameters, congestion control, fabric topology, etc. Current methods require up to 10k simulation experiments to identify optimal configurations, with inadequate exploration leading to significant degradation of training performance. In this paper, we tackle the overlooked problem of efficiently conducting parallel simulation experiments for design space exploration. Our

2024

2023

2022

2021

2020

  • [Topic] [ The path planning algorithm for multiple mobile edge servers in EdgeGO], Rong Cong, 2020-11-18

2019

2018

2017

Instructions

请使用Latest_seminar和Hist_seminar模板更新本页信息.

    • 修改时间和地点信息
    • 将当前latest seminar部分的code复制到这个页面中
    • 将{{Latest_seminar... 修改为 {{Hist_seminar...，并增加对应的日期信息|date=
    • 填入latest seminar各字段信息
    • link请务必不要留空，如果没有link则填本页地址 https://mobinets.org/index.php?title=Resource:Seminar
  • 格式说明
    • Latest_seminar:

{{Latest_seminar
|confname=
|link=
|title=
|speaker=
}}

    • Hist_seminar

{{Hist_seminar
|confname=
|link=
|title=
|speaker=
|date=
}}

Retrieved from "http://mobinets.cn/site/index.php?title=Resource:Previous_Seminars&oldid=3451"