Wu Jianfeng

SP 2 A: A basket-trading-oriented securities matching set automatic partition approach for parallel trading system

basket trading is a performance killer in a parallel matching system. However, due to the lack of effective models of representing the fundamental securities, baskets, and the relationship in between them, existing matching set partition approaches in parallel trading systems are not suitable for effective automatic partition of securities matching sets for basket trading. This paper presents a basket-trading-oriented securities matching set automatic partition approach for parallel trading system—SP2A. Base on the proposed graphical representation of securities matching tasks, SP2A firstly try to place the most related securities into the same matching set or into the least-needed ones according to the corresponding trading frequency statistics. Secondly, the rest individual securities are evenly placed among the matching sets. Workloads thus can be balanced as well. SP2A proposes an effective matching set automatic partition method for reducing the synchronous delay induced by the basket trading or securities-correlated trading. Therefore, parallel matching efficiency of such a system can be improved.

FFTP: A file-like data exchange method for high-liquid securities information based on Extended FAST

Current file-like data exchange methods of high-liquid securities information are mainly based on some high redundancy formats such as DBF, which wastes bandwidth resource and induces high disseminating latency. Besides, the file segmentation technique these methods adopt aggravates bandwidth overhead and causes extra latency. To solve the problem, this paper presents FFTP - an Extended FAST based high-liquid securities data exchange method. FFTP promotes high-liquidity in two aspects. First, by mapping records of the file-like data to FAST messages, FFTP can significantly reduce redundancy of exchange data by leveraging high efficiency encoding of Extended FAST, and thus bandwidth overhead can be reduced. Second, since the encoding procedure is based on the granularity of a record, the receiver can instantly decode records from a file segment and no retransmission of the whole file is needed, which effectively reduce latency of records of the received segments. We implemented the Extended FAST and the FFTP prototype. The experiments demonstrate while avoiding full file retransmission, FFTP can decrease DBFs size by over 70%, very close to some general-purpose compression methods such as bzip, gzip, rar, which verified the feasibility and effectiveness of the new method.