High-Level Development of Multiserver Online Games

ion: this includes automatic entity serialization and hides nearly all of the technical network communication aspects. On the other hand, RTF is significantly more flexible than reusable game engines like the Quake or Unreal engines, because it is not bound to a specific graphics engine and leaves the real-time loop implementation to the developer, who is now supported by the high-level mechanisms of RTF for entity and event handling. The multiserver capability of RTF allows to easily incorporate three different parallelization and distribution approaches and is open to be extended in future game designs. This flexible support of different parallelization concepts allows RTF to be usable for a broader range of MMOG concepts than existing multiserver middleware like the Emergent Server Engine [7] or BigWorld [8] which focus mostly on the concept of zoning. Our proposed high-level development approach is efficiently supported by the current RTF implementation, which both provides a high level of abstraction and preserves design flexibility for singleand multiserver game engines. We conducted several case studies which showed that RTF is indeed easy to use and it successfully shields the developer from the low-level tasks of online game implementation. Summarizing, RTF offers the following integrated functionalities. (1) Game data structures are specified as plain C++ entities. The serialization for the transfer over a network is done automatically by RTF, and the underlying communication implementation is optimized with delta updates to reduce the amount of data sent over the network. (2) The game logic and entities are implemented in a usual object-oriented way and are open to be integrated with state-of-the-art scripting capabilities, like LUA [9], for example. (3) The proven multiserver distribution concepts zoning, instancing, and replication, as well as their combinations, are supported by RTF. The corresponding segmentation and distribution of the game world are described on an abstract level. (4) Distribution management and parallelization of the game state processing is fully handled by RTF. Segments can be reallocated to new servers during runtime and interserver client migrations are realized in a seamless way. (5) Support for advanced monitoring and controlling capabilities simplifies the dynamic management of online games in a grid environment with resource management. Besides the in-depth evaluation of RTF’s performance characteristics, we are investigating the applicability of our approach to broader, nongame classes of applications that still exhibit a basic real-time loop structure, for example, e-learning and spatial physics simulation. Furthermore, we plan to integrate additional features into RTF in the future. In particular, additional grid-related monitoring and manageability functions are highly promising to be integrated for further enhancing RTF as a comprehensive middleware for online games.