Home Blogs How US sanctions spurred India to develop high-performance computing

How US sanctions spurred India to develop high-performance computing

PARAM is a series of supercomputers designed and assembled by the Centre for Development of Advanced Computing (C-DAC), India. The latest machine in the series is the PARAM ISHAN.

The year was 1998. India woke up to the astounding news. While Buddha had smiled in 1974, in 1998 it was Shakti that shook the desert soil of Pokhran. It shook a lot more. The US was livid. Their snooping satellites had been fooled and India had once again knocked on the doors of the Nuclear Club. MTCR sanctions followed and high technology items, called dual-use technologies, were denied to many Indian R&D laboratories, among them ISRO centres. The MTCR was applied so rigorously that a 500MB hard disc for a desktop computer was snatched off the airline loading bay and the supplier got a nasty note from the US embassy. More was to follow.

Orders for GIS software were accompanied by an undertaking that the software was not to be given or shared with several countries. A long justification of how the software was to be used and by whom was demanded. In spite of this certain module relating to Web, enablement was denied. The software had to be directly delivered to the end user (non-ISRO) even though ISRO was paying for them. A similar situation existed for workstations. One of the most painful episodes was when an order for a workstation for use with GIS matched a similar order from another ISRO centre for a workstation for structural analysis. A US team swooped down, accompanied by a very apologetic team from the Indian agent and demanded to see the physical units and the work being done on them.

The message was clear. UNIX workstations were immediately junked for off-the-shelf servers and desktops which ran UNIX on Intel Multi CPU servers as these lay outside the MTCR. Even then there were problems. ISRO had taken up the development of an Airborne Synthetic Aperture Radar. ISRO also had an agreement with ESA for the reception and dissemination of ERS 1 and 2 SAR data. The existing computer facility consisting of a mainframe with a vector processor, acquired before the sanctions, was just about to manage SAR processing but it took 12 to 18 hours to process an airborne SAR scene and 8 hours for an ERS-1 SAR scene. The National Centre for Medium Range Weather Forecasting had a Cray but also had a US-appointed gatekeeper who kept a strict watch on who came in and what was being processed, effectively ruling it out. Incidentally, CRAY is now a footnote in the history of computing in India.

At this point, two other indigenous computers were considered. One was the Flowsolver at National Aeronautical Laboratory and the other a cluster computing facility at the Bhabha Atomic Research Centre. Test runs on both these systems were promising but for the fact that SAR processing is both data and processing intensive. These clusters operated on the principle of scatter-gather, that is, assign chunks of the data to each computer and gather the analyzed results to get the final result. The inter-computer communications, in the case of SAR data, became the bottleneck in achieving the desired processing speed.

Meanwhile, the Centre for Development of Advanced Computing, C-DAC, a unit under the Department of Electronics had started working on PARAM, a parallel processing computer using transputers. The SAR requirements were of interest to them and a MoU between ISRO and CDAC resulted in a satisfactory solution. Luckily the transputer was an English device and hence not under the US embargo. In fact, India did get support from European suppliers during the sanctions regime. Using PARAM the processing time of an ERS SAR scene could be reduced from 8 hours on a VAX11/780 with a Vector processor to 40 minutes on an eight node PARAM. The entire processing system at NRSA (now NRSC) for ERS and airborne SAR was based on PARAM. PARAM was also used in VSSC for high-performance computing jobs. In time, better solutions emerged which further reduced the processing time. All these solutions did not need any hardware or software under sanctions.

Were it not for the US sanctions engineers in ISRO and CDAC would not have put in these efforts and missed out on exploring the alternative hardware and software options. In the process, they also became wiser, as did the US about the futility of sanctions in the face of a determined community.