LaGRange: Locality-aware Graph with Range Maintenance

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/95606

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/95606

Title:	LaGRange: Locality-aware Graph with Range Maintenance
Authors:	徐裕翔;Hsu, Yu-Hsiang
Contributors:	資訊工程學系
Keywords:	圖存取;圖神經網路;快閃記憶體;寫入放大;Graph storage;GNNs;Flash memory;Write amplification
Date:	2024-07-26
Issue Date:	2024-10-09 17:05:50 (UTC+8)
Publisher:	國立中央大學
Abstract:	隨著圖神經網絡（GNNs）的興起與其強大的解釋能力，對於高效存取圖結構資料的需求重新變的重要。然而圖結構資料不規則且複雜的特性加劇了寫入放大和讀取的延遲。儘管如此，現階段的針對圖儲存相關的研究主要集中在靜態圖上。雖然一些研究涉及動態圖，它們討論的是在不同的時間戳下保持該時間點下的圖，而非真正的去維護一個會隨時間而變動的單一圖資料結構。因此，本項研究提出了一個關注單一圖結構資料隨時間演變的局部感知圖快閃記憶體－LaGRange的全面設計。它能夠更有效地管理圖的更新，同時減輕寫入放大和讀取延遲。;With the rise of Graph Neural Networks (GNNs), the need for efficient graph storage has resurfaced. While retrieving data from external storage, the irregular pattern of graph data exacerbates both write amplification and read latency. Despite this, current research on graphs predominantly focuses on static graphs. Although some studies address dynamic graphs, it maintains multiple versions of a graph at different timestamps rather than a single graph evolving over time. The study proposes Locality-aware Graph SSD with Range(LaGRange) maintenance which focuses on a single graph that evolves over time. It’s a comprehensive design to manage graph updates effectively, mitigating write amplification and read latency simultaneously.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	17	View/Open

社群 sharing

Loading...