Over the last 20 years object-oriented programming languages and managed run-times like Java have been very popular because of their software engineering benefits. Despite their popularity in many application areas, they have not been considered suitable for real-time programming. Besides many other factors, one of the barriers that prevent their acceptance in the development of real-time systems is the long pause times that may arise during large object allocation. This paper examines different kinds of solutions that have been developed so far and introduces a switchable approach to large object allocation in real-time Java. A synthetic benchmark application that is developed to evaluate the effectiveness of the presented technique against other currently implemented techniques is also described.