KG SysNet

SysNetは、OS・ネットワークなど広範なシステムソフトウェアを研究・開発するグループです。

About

OSやネットワークを基盤に、カーネルなどの低層からエッジ・分散システム、データベース、Webまで、広範な領域の研究に取り組んでいます。特に、エッジ／デバイス上での推論を効率化する実行基盤の研究や、eBPFによるリアルタイム攻撃・異常検知の基盤設計に注力するメンバーが在籍しています。メンバーは関心のあるシステム領域を自由に研究でき、中澤・大越研究室のプロジェクトとの連携も選べます。

SysNetをはじめd-hacks、Sensys、WellComp等の研究室ネットワークを、運用グループ「jnroot」の中心として運用しています。ネットワークやデバイスの運用を通じて、GPUサーバーやエッジデバイス環境といった計算基盤の実運用から得た知見を研究に還元しています。

Research

PilotQ：エッジデバイス推論における確率的スケジューリングのための階層型メタ制御アーキテクチャ
Abstract エッジ推論システムは対象がスパースかつ非定常に出現する環境において計算リソースを浪費しやすいが，既存の確率的スケジューラ JumpQ は静的なパラメータ依存のため実環境での変化に追従できない．この限界は，従来定義されてこなかったミリ秒スケールの観測ノイズ，秒スケールの短期変動，分スケールの中期変動，さらには時間スケールの長期変動という，四層にわたる時間的不確実性を単一のシステム内で協調的に解決する手法が確立されていない点に起因する．これを解決するため，本研究では制御工学のタイムスケール分離原理を適用し，各時間的不確実性層ごとに専用制御ループを割り当てる階層型メタ制御アーキテクチャ PilotQ を提案する．PilotQ は入力対状態安定性を形式的に保証し，シミュレーションによって実環境変化への自律的追従による計算リソース効率の最適化を実現できることを示す．
Laqista: Serverless Cloud-Fog-Dew Computing Platform for Deep Learning Applications（IEEE SmartComp2025）
Abstract In Smart Things and smart city applications, IoT devices generate large amounts of data and deep learning technologies are used to acquire useful information from it. Based on the kind of application and data, there are various non-functional requirements, such as low latency for information presentation by MR and privacy for video processing. To serve these requirements, a computing platform needs to make appropriate use of computing resources, namely Cloud, Fog, and Dew. However, there are some technical challenges in designing such a platform: i) transparently satisfying application QoS; ii) running the application across various hardware and OSes without modification; iii) sharing the application context taking into account the validity of values (temporal locality) and the data privacy (spatial locality). In this paper, we introduce Laqista, a novel Cloud-Fog-Dew computing platform. Laqista serves applications in a serverless manner via the Edgeless API, which schedules requests and abstracts the details of the platform. Applications are separated into Logics and Models, which are converted to lightweight, platform-agnostic formats such as WebAssembly and ONNX, respectively. Additionally, the Context Store synchronizes application context among the nodes, handling the privacy and validity of data. We developed a prototype implementation of Laqista in Rust and evaluated its performance. Experimental results show that the Laqista design has practical performance and is applicable to real-time applications such as video processing and MR.
—Related Link: 【採択】卒業生 牧野君の論文がIEEE SmartComp2025にacceptされました
Dynamic Fixed-point Values in eBPF: a Case for Fully In-kernel Anomaly Detection（AINTEC 2024）
Abstract eBPF and XDP are promising technologies that are capable of accelerating packet processing inside the Linux kernel. Despite these benefits, eBPF is constrained by a number of rigorous restrictions that are imposed to protect the kernel. One such restriction is the lack of support for floating-point values, which was introduced to achieve faster execution and avoid non-deterministic behavior. However, this has become a significant obstacle to expanding the functionality of eBPF programs with advanced algorithms. In this paper, we propose dynamic fixed-point as a solution to overcome this challenge within the restrictions of eBPF. Dynamic fixed-point values are an expansion from traditional fixed-point values, with the bit allocation adjusted dynamically. Benefit of dynamic fixed-point is that the accuracy of calculations are improved, which is one of the critical shortcomings of fixed-point. To demonstrate the effectiveness of our approach, we have designed and implemented a prototype of an entropy-based traffic anomaly detection framework and have reported on its performance and the detection accuracy. Our prototype, which employs dynamic fixed-point, has achieved an 18% improvement in throughput while also matching the detection accuracy of a similar system that employs floating-point values in user space.
—Related Link: 【採択】学部3年 大崎君の研究が国際会議 AINTEC 2024にacceptされました
FaST: Accelerating Web Front-end Data Binding with Compiler and Visible Anchor（ACM The Web Conference 2024 Short Papers）
Abstract Data binding in web front-end development has made a significant contribution to removing complexity from development and simplifying programming. However, data binding has caused a degradation of website performance at the cost of reducing the burden on programmers. In this paper, we propose Visible Anchor to solve the performance degradation caused by data binding. We develop a compiler called FaST that implements the method. Then, We compared the rendering time among websites built by existing methods and FaST compiler. The evaluation result revealed that the websites built by FaST compiler are at minimum 2.9 times faster to be rendered than the ones built by the existing methods. FaST made a significant contribution to improving the performance of web front-end data binding. Consequently, data binding with FaST can be a better choice for web front-end development.
—Related Link: 【採択】修士1年 富澤君の研究が国際会議 ACM The Web Conference 2024 Short Papers Sessionにacceptされました