In order to serve industrial R&D as well as academia, three of TWAREN's four core nodes are located within Taiwan's Science Parks. Eleven regional centers (GigaPoPs), linked to the backbone as 10 Gbps Metropolitan Area Networks (MANs), serve to bring Taiwan's major institutions of higher education and research onto the network. TWAREN provides many services not available from commodity networks such as IPv6, multicast, MPLS/VPN, VOIP, e-learning, and multimedia.
The NCHC's Cyber-resources are geographically distributed across Taiwan at three different resource centers. The Hsinchu Business Unit, located in northern Taiwan, houses the NCHC's supercomputers, software library, and scientific databases. The soon-to-be-completed Taichung Business Unit, located in central Taiwan, will be the NCHC's Knowledge Management Center. It will be a Peta-scale data storage service center and serve as NCHC's Grid Operation Center as well. The Tainan Business Unit, located in southern Taiwan, serves as the NCHC's Network Operation Center (NOC). These three resource centers are linked to each other via the TWAREN network, thus allowing for the integration of all of the NCHC's computing and data Grid resources.
The Cyber-environment's major task is to build a Cyberinfrastructure on a scale that has never before been seen in Taiwan. The basic infrastructure includes a high-end computing facility and a 100 TB level storage system running on top of 10-Gbps level of Wide Area Network (WAN). Although these developments are not quite as impressive when compared to those of worldwide Grid activities, they are indeed the largest resources in Taiwan.
The build-up of the basic infrastructure is only the first step. We must also focus on how to provide stable, secure, and scalable services to our scientific computing community. Over the past two years, we have devoted the majority of our efforts to the construction of a generic computing Grid environment. This environment was built to help hide the complexity of Grid computing from the end users and to provide easy access to the backend computing facilities and storage. The fundamental concept of building such a "generic" computing Grid is to provide a base line for the development of an application-specific Grid. Application developers can easily modify this generic solution to develop their own customized computing Grid if common features are already built into it.
The generic computing Grid seems to have adopted the worldwide standard of Grid middleware, however, the bridge between the middleware and the user is not standardized. The Grid must provide special features that traditional scientific computing does not. The user should be able to simply "fire and forget." In other words, the user should be able to submit a job and then let the system take care of execution and notification of the results.
The ultimate goal of Grid computing is to provide seamless and consistent "plug and play"-style computing to its users. In order to realize this, we must enhance Grid's middleware and learn to better manage its infrastructure. Also, middleware must be adopted and integrated while new features are added. This will further reduce the barrier between the Grid and its users. Various tasks such as automation, resource brokerage, meta-scheduling, and advanced reservations can be carried out over the Grid system while, at the same time, hiding the Grid's complexity.
The large-scale problems of the future will be much greater than those of today. In the future, many modern technologies and applications will become routine and will be integrated into a much larger system that is able to solve much more complicated problems. The modern supercomputer will become no more than a simple piece of hardware that will not require further development. On the other hand, the Cyber-environment will demand intensive development in order for it to solve the large-scale problems of tomorrow. The primary tasks demanding our attention in the future will be to provide the user with a satisfying Grid experience, to further hide the complexity of the Grid computing environment and its backend facilities and, finally, to achieve the goal that Grid computing set out to achieve originally. It will not be enough to simply implement solutions that meet the research demand. We must truly transform the way research is done and inspire new means of R&D. The Cyber-environment will make all of this a reality.
