1 / 13

MSSG: A Framework for Massive-Scale Semantic Graphs

MSSG: A Framework for Massive-Scale Semantic Graphs. Timothy D. R. Hartley , Umit Catalyurek , Fusun Ozguner , Andy Yoo , Scott Kohn , Keith Henderson Dept. of Electrical and Computer Engineering, The Ohio State University Dept. of Biomedical Informatics, The Ohio State University

bambi
Download Presentation

MSSG: A Framework for Massive-Scale Semantic Graphs

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. MSSG: A Framework for Massive-Scale Semantic Graphs Timothy D. R. Hartley , UmitCatalyurek , FusunOzguner, Andy Yoo , Scott Kohn , Keith Henderson Dept. of Electrical and Computer Engineering, The Ohio State University Dept. of Biomedical Informatics, The Ohio State University Center for Applied Scientific Computing, Lawrence Livermore National Laboratory Presented by K. Sheldon April 2011

  2. Abstract This paper reports the findings from the development of a middleware framework to access and analyze massive scale graphs. The target is scale free semantic graphs with O(1012 ) vertices and edges. Since massive scale graphs do not permit storing the entire graph in memory, a new graph database, grDB, is proposed to efficiently store and retrieve large scale graphs using out of core storage. This framework scales well and the experimental results show that the grDB outperforms other databases such as BerkeleyDB and MySQL. Objectives: • Design and implement a flexible, easy-to-use API and associated middleware platform for analyzing massive-scale semantic graphs Presented by K. Sheldon April 2011

  3. Overview • Design of architecture • MSSG (Massive Scale Semantic Graphs) • Experimental results • Conclusions Presented by K. Sheldon April2011

  4. Scale Free Graphs • Follow power law • Many vertices with low degree • A few hubs have very high degree Presented by K. Sheldon April 2011

  5. MSSG Architecture

  6. MSSG Architecture

  7. grDBArchitecture

  8. Experiment • 24 nodes - dual 2.4GHz AMD Opteron 250 8 GB RAM per node 500 GB local disks • Graphs Pubmed-S: 3,751,921 vertices and 27,841,781 edges Pubmed-L: 26,676,177 vertices and 519,630,678 edges Synthesized: 100 Million vertices and 2 Billion edges • Measurement Search time (s) Aggregate Edges processed Presented by K. Sheldon April 2011

  9. Custom Software (MSSG program) • Stream graph clustering engine built at LLNL to solve the graph search problem. It can run on any cluster. Data is stored on disk. • Lessons learned: • Well designed custom software scale to billion edge graphs performs well relatively inexpensive compared to other options. Cost : long software development time lack of generality across other graph algorithm domains. Again the communication and disk writes are impediments. Presented by K. Sheldon April 2011

  10. Ingestion and Search Performance (PubMed-S)

  11. Ingestion and Search Performance (PubMed-L)

  12. Search Performance of grDB for 100M graph

  13. Conclusions • A custom disk based graph database, grDB combined with a parallel out of core BFS algorithm • Predictions: for a One trillion edge graph Ingestion with grDB in roughly 77 hours average search in 10s of minutes • I/O efficient structures are necessary with more experimental testing on performance. • Investigate other external memory libraries such as TPIE (transparent parallel I/O environment). This provides efficient access to multiple disks attached to a single system • Investigate optimization techniques for the Ingestion, GraphDB and Query portions of the framework. Presented by K. Sheldon April 2011

More Related