Concurrent High-performance Persistent Hash Table In Java

Abstract

Current trading systems must handle both high volumes of trading and large amounts of trading data. One crucial module in high-performance trading is fast storage and retrieval of large volumes of data simultaneously accessed by multiple computer traders. To speed up access, a high-performance in-memory software-cache stores the dynamic working-set of trades during a trading day. To utilize memory effeciently, it is beneficial to provide a single shared cache for multiple trading applications. Much of the cache access is read-only, as information is gathered before a transaction to determine its value. Hence, extremely fast lookup is essential to support quick information gathering for assessment. This thesis presents a software-cache, called MapHash, that is a high-performance hash-table for use in Java.