REMOTE OPERATION OF LARGE-SCALE FUSION EXPERIMENTS* D.P. Schissel, G. Abla, General Atomics, San Diego, CA 92121, USA M. Greenwald, T. Fredian, J. Stillerman, MIT, Cambridge, MA 02139, USA Abstract This paper examines the past, present, and future remote operation of large-scale fusion experiments by large, geographically dispersed teams. The fusion com- munity has considerable experience placing remote col- laboration tools in the hands of real users. Tools to remotely view operations and control selected instrumen- tation and analysis tasks were in use as early as 1992 and full remote operation of an entire tokamak experiment was demonstrated in 1996. Today’s experiments invari- ably involve a mix of local and remote researchers, with sessions routinely led from remote institutions. Currently, the National Fusion Collaboratory Project has created a FusionGrid for secure remote computations and has placed collaborative tools into operating control rooms. Looking towards the future, ITER will be the next major step in the international program. Fusion experiments put a premium on near real-time interactions with data and among members of the team and though ITER will generate more data than current experiments, the greatest challenge will be the provisioning of systems for analyz- ing, visualizing and assimilating data to support distri- buted decision making during ITER operation. INTRODUCTION For 50 years magnetic fusion energy has been an open collaborative science with worldwide participation. As the science has progressed, experimental facilities have grown in size and complexity. Concurrent with this growth has been a reduction in their number resulting in increased growth in collaborations. Today within the U.S., there are three large magnetic fusion facilities yet there are over 40 distinct organizations and ~1000 scientists engaged in fusion research. The community’s next major step is the construction and operation of ITER, a burning plasma magnetic con- finement experiment to prove the scientific viability of controlled fusion as an energy source [1]. ITER will be located in France and will run as an international collabo- ration, with researchers from China, Europe, India, Japan, Korea, Russia, and the U.S. sharing operational and scien- tific responsibilities. It is expected that the full scientific exploitation of ITER will only be possible with substan- tial and efficient remote collaboration infrastructure. Unlike very large experiments in other fields, the oper- ation of fusion energy experiments put a premium on near-real-time interactions with data and among team members to support decision making during operations. In today’s experiments, plasma pulses, or shots, are typi- cally taken 2-4 per hour with a total of several thousand per year. The average cost of a plasma pulse is large (for ITER; total project cost amortized over all pulses is ~$1M) and therefore pulses required to carry out an experimental program must be minimized. PREVIOUS WORK The U.S. magnetic fusion community has a long and successful history of placing collaboration tools in the hands of real users. As early as 1992, a major diagnostic on the Tokamak, Fusion Test Reactor (TFTR) at Princeton Plasma Physic Laboratory was being operated from an off-site location [2]. Expanding beyond just diagnostic operation, just four years later both the Alcator C-Mod (Cambridge, MA) and DIII-D (San Diego, CA) tokamaks demonstrated full remote operation from a control room setup at LLNL [3,4]. Two important lessons were learned from these activi- ties. First, the structure and efficiency of the data systems and its capability for transparent remote access were critical for successful remote operation. The MDSplus data management and data system [5] was initially devel- oped in the late 1990s and is presently in use by over 30 experimental facilities worldwide including the two new- est tokamaks EAST in China and KSTAR in South Korea. Based on a client/server model, MDSplus provides a hier- archical, self-descriptive structure for simple and complex data types and is used to store digitized, analyzed, and simulation code data. Second, it was recognized that effective tools for interpersonal communication in a geo- graphically distributed environment were crucial and needed significant improvements. These and other efforts were mostly carried out on an ad-hoc basis. To expand on these efforts, the National Fusion Collaboratory Project (2001-2006), part of the first round of SciDAC (Scientific Discovery through Advanced Computing) projects, consolidated previous work, developed new capability, and deployed collabora- tion tools to all major U.S. experimental facilities [6]. One of these new capabilities is the deployment of a national fusion energy sciences grid (FusionGrid) for se- curely sharing resources (data, codes, visualizations, com- munication) over the Internet. The goal of the FusionGrid is to allow scientists at remote sites to participate as fully in experiments and computational activities as if they were working on site. Access to data on FusionGrid uses a secure version of MDSplus that requires an x.509 iden- tity certificate for authentication. Combining this with a unique distributed authorization scheme allows stake- holders to confidently grant restricted access to data, computers or codes as required. A web-based real-time _____________________________________ *Work supported by the U.S. Department of Energy SciDAC Program and at General Atomics under Cooperative Agreement No. DE-FC02- 01ER25455 and DE-FC02-99ER54512. WPPB28 Proceedings of ICALEPCS07, Knoxville, Tennessee, USA Major Challenges 454