Download BIONIC Desktop Grid Computing Primer | CSCI 3350 and more Study Guides, Projects, Research Software Engineering in PDF only on Docsity! BOINC Desktop Grid Computing Primer *Content liberally borrowed from other sources BOINC is a platform designed originally for volunteer computing, but can manage desktop grid computing as well. It is a middleware application that manages the submission of workunits to compute nodes and retrieval of results. BOINC Components There are two major components to BOINC. The component most people are familiar with is the BOINC Manager. This is the client-side application which makes resources available to BOINC projects. The second component is the BOINC Server. This is where projects are housed, workunits created, delivered, and received. It’s also the most complex part of the BOINC system. It consists of several components mentioned below. This is where most of the effort to put together a BOINC project will be spent. Volunteer vs. Desktop Grid Computing Volunteer computing (VC) and Desktop Grid (DG) computing are very similar. The key difference is ownership of the compute nodes. If you are running a project and you neither own the compute nodes, nor compensate the owners for their computing power – congratulations, you have a volunteer computing project. On the other hand, if you do own (administratively, at least) the compute nodes, this is referred to as a desktop grid computing project. For our purposes, we’ll be creating a DG project. The difference may be subtle, but it does slightly change the workflow of the project. VC projects rely on the volunteer to specify what projects to join, how much time is donated, when the time is donated, etc. With a DG project, the project administrator supplies this information. If the administrator wants to donate 100% of available resources to his project, he/she has the ability to do so. Workflow The basic unit of work in a BOINC project is sensibly called a workunit. This is smallest measure of work recognized by BOINC. It is an administrative collection of files which includes the data to be processed, how it is to be processed, and what to do with the result. Workunits are created on the BOINC Server – presumably as smaller chunks of a larger overall project. As mentioned above the workflows between VC & DG computing are somewhat different. With a VC project the owner of the compute node requests a quantity of data for the computer to process. This is typically expressed in second of compute time. The server in response returns enough workunits to fulfill the request. The compute node processes the data and notifies the server of the results. The cycle the repeats, with the compute node requesting more work, the server fulfilling the request … ad infinitum. DG workflow is similar, but the volunteer is removed and replaced with configuration files. These configuration files are copied to the compute nodes and tell the BOINC Manager how to behave (which projects to attach to, what quantity of resources to make available, etc.) Otherwise, the process is identical.
