Soft real-time view management

Abstract

Traditional database management systems have been designed to assist in maintaining large collections of data. However, their performance is not adequate for instant communication applications with tight real-time constraints. The Italian research laboratory Telecom Italia Lab has designed and implemented SOFTRAM DB, a real-time distributed system which stores user profile data in main memory. Different real-time applications (e.g., presence services) concurrently connect to the SOFTRAM DB to store and query user profile data (e.g., nicknames, user identifiers, moods). To improve query and update performance, accessory data structures (e.g., indices, materialized views) are usually exploited in relational DBMSs. This paper presents a soft real-time view management for user profile data in SOFTRAM DB. Different main memory views (i.e., user view and application view) and structures are presented, highlighting their advantages and disadvantages. Preliminary experimental results, performed on large synthetic datasets, show the adaptability and the effectiveness of the proposed approach.