Java Peculiarities

Due to the specific way of memory management as employed by current Java virtual machines (VM), a special data handling and caching functionality is used by RTVR to support the visualization of huge data sets (dozens to hundreds of volumes), which are produced, for example, by numerical simulations [4]. The maximum amount of memory which is available to a VM has to be fixed at initialization time. As the garbage collection and object allocation mechanism sweeps through the entire address space of the VM, allocating more memory to the VM than physically available would lead to excessive paging and strong performance degradation. Instead of allocating sufficient memory to fit even the largest data sets, RTVR uses a separate memory and disk cache for space-demanding parts of it's data structures, i.e., the original volume data, the extracted voxel attributes, and the renderable voxel data. Data, which is currently not used for rendering, is placed into the memory cache and thereby potentially written to disk by a background thread. Requests for recently used data can usually be satisfied out of the memory cache, whereas reading less recently accessed data may require to fetch it from disk. The cache feature is used when large data sets are visualized locally, and is disabled when RTVR is used within a web-browser.

mailto:mroz@cg.tuwien.ac.at.