A multidatabase solution for a financial application

Abstract

We have built a multidatabase system to support a financial application that provides access to real-time and historical financial data concerning stocks, bonds, options and mutual funds. The arrival rate of data is 500 items per second, each of which may be saved in a DBMS. Sub-second response time is required for queries.The consistency and performance of real-time data differs from those of historical data. Real-time data is non-persistent but requires database consistency. Historical data is persistent and requires serializability on 〈instrument, timestamp〉 pair. Since real-time data reflects current market activity, it must be delivered to users faster than historical data. Historical data has sub-second response requirements for queries. A typical query for historical data requests between 100–1000 records.We could not find a single DBMS or even a multidatabase system that met our performance requirements and consistency definition. In this paper we define the characteristics of the application, the multidatabase system we used to support the applications, and some requirements for DBMS's in financial applications.KeywordsMutual FundFinancial ApplicationTransaction ManagementGlobal ConsistencyPersistent StoreThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.