Figure 5 - uploaded by John Shalf
Content may be subject to copyright.
Diagram of the UDP packet format. A 20-byte header was used to locate the payload of each packet in the destination domain completely independently of one-another. Assuming a full 3D block domain decomposition, the origin and dimension information is superfluous in the z-dimension. X/Y origin global => the offset in grid coordinates with respect to the global data dimensions of the origin of the domain decomposed block (local domain) in the data source. X/Y dim local => The x/y dimensions of the local domain decomposed chunk in the data source (used to control wrap-around of data as it is written into the destination data array). Localindex => The offset in cells counted from the start of the local domain decomposed block in the data source. This source-based data indexing allows the component sending the data to be ignorant of the domain decomposition at the destination, but provides enough information at the destination to support unambiguous data re-assembly.
Source publication
This past decade has seen rapid growth in the size, resolution, and complexity of Grand Challenge simulation codes. Many such problems still require interactive visualization tools to make sense of multi-terabyte data stores. Visapult is a parallel volume rendering tool that employs distributed components, latency tolerant visualization and graphic...
Context in source publication
Context 1
... the SC01 Bandwidth Challenge, we modified the back end of Visapult to work with our own custom UDP transport protocol. Each packet contains encoded information about where to place its data payload in the destination array (see figure 5). This ensures that each packet can be treated independently so that packet ordering and packet loss have minimal effect on destination processing. ...
Citations
... Bethel et al. in LBNL (Lawrence Berkeley National Laboratory) have also done lots of work in grid enabled visualization. They developed an object order visualization backend VisaPult [19] for visualization of T-bytes sized scientific data. They made the system grid enabled by connecting it to grid middleware Cactus [20]. ...
In this paper we present GVis – a Java-based software architecture for grid enabled interactive visualization. Compared with
traditional parallel solutions that use multiprocessor computers or PC clusters, GVis provides a grid supporting environment
that enables transparent conglomeration of heterogeneous resources, dynamic and autonomous coordination of visualization tasks
and collaboration among end users. A portal is provided to end user for launching tasks and viewing results. With a Java-based
object oriented visualization framework, the system can be extended and adapted conveniently to support a variety of visualization
tasks.
We present a method for enabling progressive client-server volume visualization of data from the computing grid. Rendering is performed on clients, while servers on the grid provide wavelet-compressed volume data.