Interview with S. Ramakrishnan, Former Director General, C-DAC.
- By R. Ramachandran
S. Ramakrishnan : "If anybody can do it, C-DAC should do it."
The Centre for Development of Advanced Computing (C-DAC) started out with the specific mandate of developing a High Performance Computing (HPC) system and providing HPC solutions for various applications. Today C-DAC has diversified into many other information technology-related activities and has also expanded its portfolio of institutions. C-DAC's Former Director General S. Ramakrishnan spoke at length to Frontline about the organisation, its achievements and its goals.
Do you see C-DAC's diversification into other IT-related activities and expansion of its portfolio of institutions as a positive development for the future of the organisation?
I would not like to look at it as positive or negative. I look at it more from a more pragmatic point of view. Companies that have stuck to one product line quite often perish simply because they have been blown over by events. With regard to both the bouquet of topics as well as institutions, some of them are due to certain accidental developments and some of them are part of a natural organic growth process. After C-DAC built the supercomputer PARAM 8000 as a product, we had to move on. People asked the next question "whom does it serve?". Then, one after another you will find that from a product you move on to solutions for a diversity of users, each one having a specific need. This effectively also meant associated training or capacity-building of users for applications.
In a sense, you had to create a whole ecosystem around HPC. That's why even within HPC we ended up as a one-stop shop, ranging from setting up HPC facilities and building application groups. We also entered allied areas like the new dimension of problem-solving through distributed computing using a grid. Another side is how you take on other topics, say for instance, language technologies. The story of how language technologies got added is like this: By 1988-89, C-DAC had begun its operations and had a critical mass of people and an exciting spirit was there. Here was an area that was created through department [Department of Electronics and Information Technology] funding and incubated at IIT [Indian Institute of Technology] Kanpur up to a stage when it needed a technology house where it could be taken further and finally to the market. That's how both the creator and the basic technology came here. Then it exploded as a whole new paradigm in terms of the GIST [Graphics and Intelligence based Script Technology] group and the language products and related things. Therefore, one way of looking at C-DAC [is] first as a product and a technology house that became more of an idea house. As long as you have a creative place where people enjoy and do things, and if you can maintain that, not restricting yourself to this or that per se need not be a problem.
Beyond that, multiple institutions got merged together. At least in the last five to six years, we have viewed the amalgamation as healthy rather than conflicting. In a way, we have become what we are. The challenge is to maintain the early spirit of a place where people enjoy. The outcome will then automatically turn out to be exciting.
When C-DAC started on HPC, you had to do every aspect of it yourself: software, hardware and architecture. Today much of the hardware is available off the shelf. The focus now is more on the architecture to obtain highcomputing speeds. What would you say are C-DAC's niche areas in HPC in recent times?
Today there is suddenly a new surge on multi-core processors. What was happening in parallel computing [at the system level] is now happening at the board level, node level, etc. So, we have to do a lot of parallel multi-threading, etc. In the context of parallel programming, this itself is going to be an interesting topic that is going to dominate in the next five to 10 years, that one has to master. Second is the overall petaflops [1,015 floating operations a second] capability and beyond, which are definitely going to be hybrid architectures. In supercomputers, by their very nature, there is a relationship between applications and the underlying architecture. You have to keep playing around [with the architecture] all the time. How do you make the systems perform best ? That's where algorithms come into the picture. And then, of course, with hybrids, accelerator boards and accelerator cards also come into the picture.
So as you go forward, at each next generation, even though hardware is available, it needs to be tested. Then power is going to be a serious problem. So we will be revisiting old issues. Now people are talking of "green" computing systems because we are pushing ourselves and consuming too much power. So, in a sense, the more you go forward, the more you come back to some of the old issues. So one asked the question, instead of one single "Babel Tower", why not have multiple smaller systems distributed at multiple locations and loosely coupled so that we consume less power. But that does not really solve some classes of problems as well as a single high-power system. So today's challenge to C-DAC or any other player in HPC would be to address these issues. You break down the given architecture into pieces and address each one of them.
There are many groups doing similar things in the country today. But having evolved in the background of HPC, where does C-DAC's competitive advantage lie today and what are the new challenges?
At the time when C-DAC started there was Cray [the American supercomputer company] and we were denied that technology. That defined the starting problem for C-DAC. Having proved a point, you saw in the 10 to 15 years of reforms, the whole paradigm shifted for those who were doing import substitution. So we also need to look at areas where we do not have a comparative advantage because we have to tackle that also; whether it is products or services. Whether it has to be technology of products or IP [intellectual property]. These kinds of debates are always there. So in supercomputing, where does C-DAC stand and who are our choices? You know what happened to Cray. They themselves have to justify the volumes problem versus very customised highly expensive hardware machines. Everybody has gone through these challenges as the industry itself has changed. So from C-DAC's point of view, the main challenge is how you can do reasonably good volumes. In some sense, we are not producing enough systems out there in the market. That is one of the biggest challenges. For that we have had some ideas. We have been working on institutional mechanisms, how to push technology and products into some suitable instruments, etc. The reason is, at the end of the day, economics matter.
From an Indian point of view, the area where we can still do well is R&D. We can be very competitive in services and in applications, particularly where we can marshal our partnerships with various institutions. So, if you play these to your strength and bridge the gap, you don't necessarily do it all yourself, but you can partner with people. Keeping these in view, we have put PARAM Yuva up. That was the point where we actually pushed our interconnect technology to somewhat contemporary levels. But the key thing again is can we keep the pace ?
So you clearly see a role for C-DAC's continued R&D in HPC even though there are a multitude of players in the country today and also a lot of HPC systems are available off the shelf.
The answer is very clearly yes. We do see our position in the user community and in the market as very strong. And you get the benefit of all the skill sets being under one roof. Also, while I mentioned economics being unfavourable in one area, it is very favourable in another area. In terms of expertise, value for money, we are still very competitive. In terms of offering third-party facilities, only we have been offering it. This helps us bootstrap many things. And we have done it very well and exploring partnerships with application developers therein. C-DAC's role was one-dimensional in 1988. We strongly feel that it is multidimensional today.
You mentioned the problem of volumes. Do you have some idea of how to tackle this, particularly in making greater inroads into the Indian market ?
I think in a very simple way. Look at the large number of institutions existing and getting created in the academia and research. Also, look at the large amount of, what you may call, hand-holding they all need. Very recently, we set up a facility in the North-Eastern Hill University [NEHU, Shillong]. It was set up in a period of 20 days. It has all the technology components that we had to offer. Promptly came the request for a joint workshop. Can you not do more application development for the entire north-east ? So, just buying the latest machine and giving it to somebody is not necessarily the only part of the solution. There is much more than that - from technical workshops to applications development to capacity building to basic systems supplied to systems engineering to putting our technology components and technology proofing of a kind. So, if you look at the number of institutions out there in the country that are not even into the HPC culture, all the way up to setting up HPC labs for them and providing education and training, we have an exciting agenda ahead of us. In terms of volume, the answer is there to see just by listing the number of potential users.
As you diversified, you have also pioneered in several areas in the country such as language technologies, e-governance and grid computing. Could you highlight some of C-DAC's key achievements in areas other than HPC ?
In terms of the number of years of leadership and sustenance, language is very clearly [a major achievement]. C-DAC's contribution to Indian language computing has been really significant. From the foundational work ever since the early days, now that particularly many more institutions have got merged, it has become one single institution that does research [in the field], brings out the basic tools, the resources, the corpora and makes these things widely available, and also initiating constant work on adding to the spell checkers, etc. in all the 22 [Indian] languages. Once you say you have done something, that's not the end of the day. Tomorrow you need search engines in those languages. And research in speech-to-text, text-to-speech and speech technologies in all the languages. Then you have OCR [Optical Character Recognition], OHR [Optical Handwriting Recognition] and cross-lingual information retrieval. In all these projects, we are working jointly with the community. In the area of software, we are carrying on from the great tradition of NCST [National Centre for Software Technology, Mumbai]. In the e-governance space, we have brought out products, which are getting rolled out. In open source software, we have bundled the Indian distribution of desktop operating system, called BOSS, because classically it is "we can use" kind of a situation but now it has become a rallying point. The distribution is working out very well and it is coming out very strongly. In the case of grid computing, the last three to four years of investment is coming out in a very nice way. Recently, we inaugurated the Indian Grid Certification Authority (IGCA). So, in terms of moving applications to the next generation of production and day-to-day use, C-DAC has come to a very mature stage. There are other areas also but the key point is we are maturing one by one in other areas.
Is there any area that is unique in character ?
We started about four years back on quite a number of initiatives on ICT [information and communications technology] for development. We did good work on products that have good applications in rural settings. We have a centre in Bangalore for this. Some two years back we did some work on the content side like agriculture, primary education, health, rural energy, e-governance, etc., in all the Indian languages. We partnered with 70 institutions. Some of the interesting areas that C-DAC has moved into include, on the one side, building institutions which are secure through cyber security and building infrastructure like professional electronics-based power electronics display for the industry and, on the other, touching everyone's life. Infrastructure is getting rolled out in telecom and things like that, but at the end of the day, there is much more to be done on where to use these. If 100 million desktops are used, 500 million mobiles are used and a good number of them take advantage of Indian languages - be it text or be it voice - if C-DAC brings significant value to it, it would touch all people. We always tell ourselves if anybody can do it, C-DAC should do it. We have been working for the last three to four years in that direction and we are seeing some of it getting integrated into the national landscape.
The problem of volumes that you mentioned with regard to HPC systems perhaps afflicts other C-DAC products as well. Do you see it as a limitation that C-DAC, essentially an R&D organisation, is also required to provide products, services and complete solutions ? Do you think some other mechanism such as spawning a company could work better ?
This is definitely inevitable. Simply because you won't find any case of an institution started out with R&D moorings continuing by itself to deliver solutions and with a scalability that is endless. Because, either we will run out of gas or we will have to limit ourselves to small footprints. So, we think if technology is our USP that distinguishes us from other market players, we should perhaps excel ourselves in that. Of course, we have also to answer "technology for whom?" That's how we have gone ahead in the past and taken it to the user. That does not mean that we only have to take it to every user. That's how in many cases C-DAC has done a better job of converting technology into a product and made it easy for us or anybody else to take it as a service to the end-user rather than offering each one as a service. As regards how we will go about it in the future, there have been some serious thoughts as far as the market side is concerned, beyond R&D and product development, if you want to scale up to reach a large clientele or a large market share. The idea of spawning a company has been under discussion and it continues to be under discussion. I guess we will be working actively with the government on this in the near future to come to a fruitful conclusion.
In terms of providing solutions, products and services to the government sector's special needs, what are C-DAC's unique capabilities in comparison with other players in the market ?
If we are regarded well in government as a good technology player, which I like to believe is so, these are all good positive impressions simply because in terms of good sound technology, with commitment at the same time to delivering solutions, products and services and willing to stand by it, the combination is not too common. And if that goes by the C-DAC branding, it's perfectly fine. One of the major plus points of C-DAC is that it is one of the few institutions that are always ready to collaborate, whether it is public-private partnership or it is jointly with the user or it is with some academic institution like IIT.
Having willy-nilly acquired many institutions, the other asset that C-DAC has is its diverse human resource strength. How best are you able to leverage this resource towards a wider R&D skill base and enter new are as ?
As I said earlier, I see this positively as a much wider bouquet of skill sets. Second is the larger geographic footprint. We are now located at Kolkata [in the east] to Mohali and Noida [in the north] to Thiruvananthapuram down south and in other places. So, it is fairly well spread. For instance, our grid computing has been able to leverage on that. So many other projects have been able to leverage on our physical presence in multiple locations.
Almost always users want solutions. They don't know whether it is hardware or software and they want a multilingual interface for many applications. So, the applications group banks on the language group to make it a multilingual solution. You find a lot of interrelationships.
Health informatics came around as a result of multiple issues. E-governance is also leveraging much more on that. And the middleware between the software group and the HPC group has a lot of commonality and they have leveraged on that. One more illustrative case is the new topic of ubiquitous computing or pervasive computing that started two years back. It is very much in the future and it will spawn out in many forms - context sensitive computing, wireless sensor network - basically, where the physical world merges with the computing world.
So, you find all devices around you, sensors around you suddenly integrate with us in the computing world. That topic itself leveraged on the embedded systems group and some people working on sensors and wireless on the one side and people working on middleware and software aspects of it on the other. And another set of hardware people are willing to build the device, all the way from VLSI [very large scale integration] chip set design onwards. So we do see many more topics beginning and we have to keep watching all the time. Only thing is we have to be careful in choosing what would become viable.
At the end of the day, C-DAC has to be very conscious about optimising resources. But that does not limit how much people can learn, how much we can tap internally from multiple disciplines. From that perspective, having a large pool is a great asset.
What is the financial status of C-DAC now ?
I would say that in the last couple of years the government has been very generous to us. Surprisingly, we work partly in an entrepreneurial way, in the sense that various groups and various centres are encouraged not to live completely on grant-in-aid kind of style of doing things. At the same time, the government has been very supportive of many important initiatives like open source activities and grid computing. All these we could initiate only because the government was willing to see the merit of that idea and fund it. Having done that, we do encourage on the overall that we generate at least a good percentage of our budget ourselves. Today, we are able to make ends meet but we are also growing. Therefore, I would think we will need a little support.
How do you see the future ? Would you be able to become self-sustaining at some point ?
As I see it, the future of C-DAC is very bright. I won't oversimplify on the question of standing on one's own legs. Because if you just simplify that line of thinking, then why should C-DAC be in the government ? You know the current debate on the role of public institutions. So, one has to see clearly what the value that it brings to the table is. I see it as a two-track approach. One is from the point of view of entrepreneurship in the sense what new ideas it creates and engineers practical solutions and technologies. Another barometer is from the point of view of public good. Public good can be from the point of view of the States, for example language computing. Multinationals will not work on that. We think it is important for us to do that.
In the early phase of innovation, it costs you money and you have to keep doing that. But once the cherry-picking stage comes, lots of people will enter into it. Even C-DAC will not be the full gainer at that stage but it will still be contributing to other people benefiting from it.
So I would not unduly benchmark purely on self-sustainability per se but slightly keeping that [in view] helps us to be on our toes. Whichever way you look at it, I see a very bright future for C-DAC simply because there is a vibrant energy in the organisation and people see exciting work ahead of us in a lot of topics.