High Throughput Architecture Design For Real-time Market Data Dissemination Applications

As the development of financial market and IT technology, more and more financial trading and market monitoring operations are moved to the internet. One significant character of these electronic market data dissemination applications is "real-time", because the investors should get the real-time market data and analyze their investing portfolio. Usually, these financial applications often deal with a huge volume of data and their user number is also large. The "real-time" and "a huge volume of data" ask these applications with high throughput. The object of this paper is to design architectures with high throughput for these real-time trading data dissemination applications.The two main works of this paper:1) While designing these architectures with high throughput for these real-time trading data dissemination applications, the architects can reference the following strategies and methods: 1. multi-tier architectures; 2. add Cache module; 3. adopt the delta data mode; 4. adopt the sparse data mode; 5. Client polling vs Server push for real-time trading data; 6. Structuring data for performance considerations. At the same time, the architects should design the architects of highest throughput within the cost constraints and future migration cost.2) The above strategies and methods are applied to a specific real-time trading data dissemination application RTM which is a web application. It not only provides its users the real-time market data, but also some market reporting, statistic and comparison information. In the client side, these data will be displayed mainly in charts. Rich Internet Application (RIA) is adopted for RTM client side development.There are two directions for RTM extensions to improve its throughput: 1. Vertical extension; 2. Parallel extension. In vertical direction, Flex supports four data access manners: Web Service, Http Service, Remote Object (AMF) and Flex Message Service (Server push). In parallel direction: 1. Single instance of RTM server; 2. Multi-instances of RTM server without load balance; 3. Multi-instances of RTM server with load balance.