A Novel Cache and SSD-based Index Structure for Health Record Indexing

Abstract

Textual-based indexing has been widely used in various medical databases. It plays increasingly important role in modern medical information systems since it directly affects the diagnosis time. Building an efficient indexing system will greatly improve the quality of treatment and save lives in emergency departments of hospitals. Researchers have set up various algorithms to optimize the query response time. B+ tree is ubiquitous in all kinds of databases as a well-known index structure with superior performance [1]. However, it shows poor performance especially when the database operation involves many retrievals. It has been proven that query performance can be significantly improved if the node size is equal to cache size. Fractal prefetching B+ Tree was developed by a group of researchers to improve the retrieval performance. Cache-first approach produces good results but it cannot simultaneously utilize CPU cache and SSD page. We propose a multi-dimensional tree structure named CacheSSD-Based tree (Fig.1).