3608

A High-Performance Multi-user Service System for Financial Analytics Based on Web Service and GPU Computation

Gow-Hsing King, Zong-You Cai, Yan-Ying Lu, Jan-Jan Wu, Hung-Pin Shih, Chao-Rui Chang
Dept. Inf. Manage., Van-Nung Univ., Taoyuan, Taiwan
International Symposium on Parallel and Distributed Processing with Applications (ISPA), 2010

@conference{king2010high,

   title={A High-Performance Multi-user Service System for Financial Analytics Based on Web Service and GPU Computation},

   author={King, G.H. and Cai, Z.Y. and Lu, Y.Y. and Wu, J.J. and Shih, H.P. and Chang, C.R.},

   booktitle={International Symposium on Parallel and Distributed Processing with Applications},

   pages={327–333},

   year={2010},

   organization={IEEE}

}

Source Source   

643

views

In finance, securities, such as stocks, funds, warrants and bonds, are actively traded in financial markets. Abundance of market data and accurate pricing of a security can help the practitioners arbitrage or hedge their position. It can also help researhers and traders design better trading strategies. In this work, we develop a pricing and data/information service system for financial analytics with the following goals: (1) supporting fast pricing and data/information services for massive multiple users, (2) saving cost in hardware equipments and reducing energy consumption, and (3) easy maintenance and service expansion of the system. To achieve the first two goals, we use one traditional server paired with one Tesla C1060 and a set of PCs each paired with an Nvidia x275, and enhance GPU performance using a set of simple yet very effective optimization techniques. For the third goal, we develop our service system based on Web Service and the Service-Oriented-Architecture design principles. Our initial experiment results show that our GPU-based service system can deliver 4.8 tera flops computing speed, which achieves over 6000% performance increase compared to a cluster of eight Intel i7 quad-core servers. Cost-wise, the GPU-based service system costs 40% of the i7 server cluster, and consumes 50% of energy that is required by the i7 cluster. We also gives an overview of our system architecture and describe the workflow for processing pricing and information service requests.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2017 hgpu.org

All rights belong to the respective authors

Contact us: