Keywords: Parallel Computing, Clustering, Speedup, Remote Sensing
Abstract
There are a growing umber of people who want to use remotely sensed data and GIS data. What is needed is a large-scale processing and storage system that provides high bandwidth at low cost. Scalable computing clusters, ranging from a cluster of (homogeneous or heterogeneous) PCs or workstations, to SMPs, are rapidly becoming the standard platforms for high-performance and large-scale computing. To utilize the resources of a parallel computer, a problem had to be algorithmically expressed as comprising a set of concurrently executing sub-problems or tasks. To utilize the parallelism of cluster of SMPs, we present the basic programming techniques by using PVM to implement a message-passing program. The matrix multiplication a d parallel ray tracing problems are illustrated and the experiments are also demonstrated on our Linux SMPs cluster. The experimental results show that our Linux/PVM cluster can achieve high speedups for applications.
1 Introduction
There are a growing umber of people who want to use remotely sensed data and GIS data. The different applications that they want to required increasing amounts of spatial, temporal, and spectral resolution. Some users, for example, are satisfied with a single image a day, while others require many images a hour. The ROCSAT-2 is the second space program initiated by National Space Program Once (NSPO) of National Science Council (NSC), the Republic of China. The ROCSAT-2 Satellite is a three-axis stabilized satellite to be launched by a small expendable launch vehicle into a sun-synchronous orbit. The primary goals of this mission are remote sensing applications for natural disaster evaluation, agriculture application, urban planning, environmental monitoring, and ocean surveillance over Taiwan area and its surrounding oceans.
The Image Processing System (IPS) refers to the Contractor-furnished hardware and software that provide the full capabilities for the reception, archival, cataloging, user query, and processing of the remote sensing image data. The IPS will be used to receive, process, and archive the bit sync remote sensing image data from the X-band Antenna System (XAS) of NSPO. The XAS is dedicated for receiving the high-rate link of the earth remote se sing data from ROCSAT-2 satellite, and has the capability of receiving down link data rate up to 320Mbps. It will also be expanded to receive data from other remote sensing satellites. Generally, IPS has configuration to receive satellite data like that depicted in Figure 1.Remote sensing data comes to the IPS via either a satellite link or some other high-speed network and is placed in to mass storage. Users can the process the data through some of interface.
What is needed is a large-scale processing and storage system that provides high bandwidth at low cost. Scalable distributed memory systems and massively parallel processors generally do not fit the latter criterion. A cluster is type of parallel and distributed processing system, which consists of a collection of interconnected stand-alone computers working together as a single, integrated computing resource [1,2,3] .A computer ode can be a single or multiprocessor system (PCs, workstations, or (SMPs) with memory, I/O facilities, and a operating system. A cluster generally refers to two or more computers (nodes) connected together. The nodes can exist if a single cabinet or be physical separated and connected via a LAN. A interconnected (LAN-based) cluster of computers can appear as a single system to users a d applications. Cluster nodes work collectively as a single computing resource and fill the conventional role of using each node as a independent machine. A cluster computing system is a compromise between a massively parallel processing system and a distributed system. A MPP system node typically cannot serve as a standalone
Figure 1: System Architecture of IPS
computer; a cluster node usually contains its now disk and complete operating systems, and therefore, also can handle interactive jobs. I a distributed system, nodes can serve only as individual resources while a cluster presents a single system to the user.
I recent years, the performance of commodity-off-the-shelf (COTS) components, such as processor, memory, hard disk, and networking technology, has improved tremendously. Free operating systems, such as Linux and Free-BSD, are available and well supported. Several industry-standard parallel programming environments, such as PVM [ 5 ] ,MPI, and Open MP, are also available for, and are well-suited to, building clusters at considerably lower costs. Such a system can provide a cost-effective way to gain features and benefits (fast and reliable services) that have historically been found only no more expensive proprietary shared memory systems. The main attractiveness of such system is that they are built using affordable, low-cost, commodity hardware (such as Pentium PCs), fast LAN such as Myrinet, and standard software computers such as MPI, and PVM parallel programming environments. These systems are scalable, i.e., they can be tuned to available budget and computational needs and allow efficient execution of both demanding sequential and parallel applications.
RSS (Ground System Section) currently operates and maintains an experimental Linux SMP cluster (SMP PC machines running the Linux operating system) which is available as a computing resource for test users [ 6 ] . The Ground ' s Linux SMP cluster (LSC) machines are operated as a unit, sharing networking file servers, and other peripherals. This cluster is used to run both serial and parallel jobs. I this paper, an experimental environment for remote sensing and telemetry application development on a cluster is proposed. The cluster would provide a mechanism for the scientist or engineer to utilize high-performance computer systems without requiring extensive programming knowledge [ 7 ] .
As the growing volume of satellite data increases with the growing umber of users who want to process the data, there is a need to move away from the traditional computer to more powerful supercomputers. The cost, however, of these computers generally places a constraint o the types of users. Clusters of parallel computers provide a good ratio of cost-to-performance and it is within this framework that we design LSC.LSC is aimed at making it easy to process large amounts of satellite data quickly. This is accomplished through an environment tailed towards design of parallel component-based software that can easily be connected and gain high-performance.
