Abstract
dc:description.abstractThis work presents a hardware prefetcher to improve the performance of accessing graph data representing large and complex networks. We represent complex networks as graphs, and queries amount to traversals on the graph. Unlike conventional memory hierarchies that exploit spatial and temporal locality, we observe that graph traversals do not necessarily exhibit these same notions of locality. This results in degraded performance of the memory hierarchy. Consequently, our hardware prefetcher exploits locality that is intrinsic to graph traversals, which we call graph-locality to improve the performance of the memory hierarchy. We design and evaluate our prototype using a micro-architectural simulator, and deploy benchmarks from GDBench that is oriented to evaluate the performance of graph database systems.
Degree
thesis:*- Grantor dc:publisher
- University of Waterloo
- Year dc:date.issued
- 2015
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Yin, Zhuoran
Subjects
dc:subject × 3Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- http://hdl.handle.net/10012/9291