Spider II: A component-based distributed computing system

Spider 11 is the second version implementation of the Spider project [9] in the Department of Computer Science at CSUSB. Spider 11 is a distributed virtual machine running on top of Unix or Linux operating system. Spider 11 features multi-tasking, load balancing and fault tolerance, which optimizes the performance and stability of the system. Spider II system consists of four major components: Task Manager, Registry Server, Object Service Broker, and Object-Servers. The Task Manager manages the Object-Server's usage. The Registry Server detects available Object-Servers. Object Service Broker controls the data communication and distributes the computation tasks and data to the Object-Servers. Object-Servers are the workstations on a network actually doing the computation. Three distributed applications are developed in order to test the performance of Spider 11: distributed matrix multiplication, distributed prime number search, and distributed Quick Sort. The tests show that for the first two applications, Spider 11 running on a group of low-end workstations on a network has better performance than a single high-end server. For the third application, Spider is slower but the increase in computational time for sorting very large number of data items is less than in a single high-end server. Overall, we have achieved to build a distributed system that is entirely component-based for ease of maintenance.