• Twitter
  • Facebook
  • Google+
  • LinkedIn

Building Sellable Products in Academic Labs: We embarked on a journey not often followed in India

Ashwin Gumaste demonstrating indigenously developed router to the Indian Prime Minister on his visit to IIT Bombay

Faculty Interview : Ashwin Gumaste, Institute Chair Professor, Department of Computer Science & Engineering, IIT Bombay

Congratulations on receiving the Shanti Swarup Bhatnagar Prize for Engineering Sciences 2018, one of the highest Indian science/technology awards. This recognition is just another jewel added to your crown with honors such as Swaranajayanti fellowship of the Department of Science and Technology (2013), P.K. Patwardhan Award (2013), Vikram Sarabhai Research Award (2012), IBM Faculty Award (2012), Telecom Center of Excellence Award from the Ministry of Communications and Information Technology (2011), Fellow, Indian National Academy of Engineering (INAE) (2018), the Young Engineer Award of the Indian National Academy of Engineers (2010) and many others .

We are honored to have you as our professor.

1. How do you feel about receiving these awards & recognitions? What changes did these awards bring in your life and profession?

[Ashwin]: The awards, especially over the last 10 years have been reasonably satisfying and humbling. It is a surreal feeling at time. Having said that, at a meta level, the awards do not mean much. Each day is a new day, each paper submitted, is a new paper, goes through a new set of peer-review, each product has to work first, and sell next. And all these have no relation to the success/awards, etc. of the past. You are only as good as your last success.

2. You have made outstanding contributions in the field of Telecommunication networks. 180 conference and journal papers are published in leading conferences and journals, 26 US patents granted and still more in pipeline. What is your motivation that keeps you going?

[Ashwin]: I am trying to enjoy the journey of research. All these numbers, like patents, papers, etc. are part of a joyous research/development process. So far, by grace of Ganpati, the process has been nice, fulfilling and enriching. There have been much more failures than success. It takes some time to understand that one has to enjoy the failure, but after that point, one tends to enjoy constructive criticism by reviewers, market forces, colleagues, etc. The failures are the ones that make us who we are. The motivation was to do something for technology, for the country, and more importantly for the group. To shape the eager young minds for a better tomorrow, their own, and for the country’s. We had a strong product development program, which churned out good indigenous technology. I do hope that more people choose core engineering as careers than just a step towards a career in consulting/IT/finance.

3. The Carrier Ethernet Switch Router (CESR) developed by you was transferred to PSU ECIL. It is the largest technology transfer between academia and industry in India till date. Could you please explain this technology? Are there any challenges in this field that still needs to be addressed?

[Ashwin]: Networks work on the principles of layers, and each layer is abstracted out to reality with multiple protocols, often these protocols are redundant, and even when they are not, the novelty that one protocol has over another is limited. So, why not just build a single protocol that can work across all the layers of networking, at least within a closed domain? That is when we came up with the idea of Omnipresent Ethernet, which modified the hugely success Ethernet standard to make a generic switching, transport, routing and forwarding mechanism. Omnipresent Ethernet, or OE, worked on data-plane and control plane segregation, with all the intelligence and smartness relegated to a centralized controller, while the switch-routers would essentially be just forwarding elements. Even with the reduced complexity, we had to ensure that a million plus packets could be processed/forwarded in less than a second, or a single packet could be forwarded through our device in less than 1 microsecond. We did not have access to switch ASICs, and even if we would have got an ASIC, we would have had to play with its SDK to make it work for this new spec that we had designed. Hence, we chose to work with FPGAs, or Field Programmable gate arrays, which are essentially programmable ASICs. Getting the FPGA to work with 100k+ lines of RTL code, few million gates was a challenge; getting the controller to work was also not easy. Another challenge was to design a management software that could control a network of these boxes. The software had to be intuitive, as well as have enough network automata to give desirable results and feedback in real-time for customers to deploy revenue-bearing services. We designed and developed three boxes: a core network platform for large enterprises and city-wide providers, an edge platform for data-centers and a table-top box for home-gateway, small business type applications. The salient feature of these platforms was that even at layer 3 forwarding (IPv4, IPv6, segment routing) we could do 1-microsecond port-to-port latency. In fact, the latency was so good that some colleagues insisted that the platform was a switch and not a router, but today 9 years later when you see the deployments, they are all in routing domains.

Our technology was a precursor to the SDN (Software Defined Networking) paradigm. In the future, products are likely to be much more-smarter. Right now, switches, routers, firewalls, middleboxes are all dumb forwarding or rule-based action-oriented elements, but it is now a question of time before these boxes become intelligent and handle messages/packets in a smart way using tools like network analytics and deep learning. We already have designed and filed a patent for a router that does not use any forwarding protocol, but simply uses network traffic to figure out and to learn what to do with an incoming packet. This table-less router is based on a Graph Neural Network, and I am hopeful, it will be the start of a new area of research in networking. These are early days but there is going to be a good churn of such technology in the future.

4. What was the most exciting thing you came across while developing a data-centre technology for a tier-1 provider (MTNL) in Mumbai?

[Ashwin]: Making a product is one thing, deploying it is another. We were blessed to have MTNL as our first customer for the router we made. We have installed a large number of routers in various data-centers, and not a single router has failed in 9 years of operations. Those data-centers run traffic (services) for financial brokers, surveillance systems and others, where latency is key, and security is of utmost importance. Getting such a project was exciting – the Ministry of Electronics and Information Technology helped us significantly. We got the data-center certified for Tier 3 certification initially. The board of MTNL helped quite a bit. We were given a free-hand in the data-center design. Thereafter we installed the routers in a city-wide network. We wrote a paper on the experiences of the data-center as well as a city-wide network that deployed our routers. The paper won a best paper award, so we had the dual satisfaction of a happy customer and academic acceptability.

5. Can you please share some specialty of 1Tbps SDN white box which you developed for the Ministry of Defense? How does it feel to be able to contribute to your nation in this manner?

[Ashwin]: This project was initiated under the auspices of a distinguished alumnus of the Institute, the late Shri. Manohar Parrikar ji, who was at that time the Defense Minister of India. The late Shri. Parrikar ji asked us in 2016 to build a unique product for the Indian armed forces. This led to a signing of an MOU in 2016 between IITB and DRDO to develop the Transport Cross Connect (TCC). A video on the TCC and our earlier routers is on our website (www.cse.iitb.ac.in/~gnl/video.mp4). Making a large router from just FPGAs and ensuring that the end product is secure and commercially viable was not easy. The design had to be water tight in the sense that putting in so much of logic (almost 100K lines of RTL code per chip, and 16 such chips in the system) meant significant iterations. The PCB itself was quite complex considering that we were running 280 Gbps per slot in a compact 8-slot system. The management software has to interact with the hardware. Making all this work and make it available in a war-time environment was a challenge. We then transferred the technology to DRDO. We do hope that they make good use of it. Finally, customer satisfaction is what we all aim for isn’t it?

6. You worked on so many technologies that have direct applications to current technological needs of the society. Which project did you think in your career was the most challenging one so far?

[Ashwin]: All projects are challenging. If not, then why do them ?. The router development effort was certainly challenging. Building sellable products in academic labs is not easy. We were embarking a journey that was not that often done in India.Having said that, the most challenging project was the development of our first router. Apart from the development projects, two other projects which were challenging were (a) network sanitization and (b) helping a service provider change from a circuit-based network to a packet based network. The first project, was to find out trapdoors in certain type of equipment. This was a difficult task as we had to go into the hardware and software architecture of a third-party product. Post a laborious three-year effort we were able to find trapdoors in these equipment, which was very satisfying. It is a pity that much of the banking and financial networks continue to be designed using this equipment. The second project was for a major private telco. After about a year, and much resistance from the hierarchy in the telco, we gave them a roadmap to redesign their network. We were able to analytically show that the telco had potential to make 3-5x the revenue at 1/3x the manpower with the technology upgrade that we suggested. The telco resisted, and limped and it is today on the verge of yet another year in the red. I hear that finally they are now following some parts of our initial technology roadmap.

7. Could you please tell something about your latest journal paper “VNF Availability and SFC Sizing Model for Service Provider Networks” published by IEEE Access Journal? what does it emphasize on?

[Ashwin]: This paper discusses ways to grow a network which has multiple data centers to include the new paradigm of service chains or network function virtualization (NFV). The idea is to deploy scalable software on servers to do the job of what was previously done by dedicated hardware. The design problem is a constrained optimization problem and since we considered availability as one of the key goals of our problem, the problem became non-linear (availability is often computed in series/parallel connections, with the parallel connections making the computation non-linear). To this end, we proposed a neat linear approximation, which gave quick and tractable results. Our approximation was applied on a large provider network, and we have shown sizable cost savings compared to many contemporary approaches.

8. Please enlighten us about your patent FISSION: Flexible Interconnection of Scalable Systems Integrated using Optical Networks for Data Centers.

[Ashwin]: The FISSION data center patent is an idea that can be applied to large data-centers. Data-centers are the brain of the Internet, and store data as well as compute machinery (servers, VMs). It is desired that we have large data-centers. But there is a problem, connecting servers to each other inside a data-center requires a switch, and a switch has only so many ports. Even combinations of switches, in typical leaf-and-spine architecture is scalable only to a point, after which performance degrades. The FISSION architecture uses instead of an electronic switch, an optical bus, in which multiple servers can listen when a server speaks, and each server has its own set of frequencies (wavelengths) on which it speaks. Using this kind of an interconnection pattern, FISSION claims of being able to scale to very large server/node counts to the tune of even a million servers with full bisection bandwidth. FISSION uses off-the-shelf optics to create this optical backplane. We have actually built a small FISSION prototype. We also have simulated FISSION and have shown its working for a million nodes. Recently we were awarded two patents by the USPTO in less than 2 years of filing on the architecture and system reliability of the FISSION paradigm. Aniruddha Kushwaha (this years’ Ph.D. dissertation awardee) held the Google student fellowship – the first such for IITB, for his work on FISSION architecture.

9. What are some notable researches done by your colleagues that you have admiration for? Is there any other research work that you admire and you’d like to explore in future?

[Ashwin]: There are some really great people at IIT Bombay. The people who inspire me include, Vikram Gadre (EE) (signal processing), retired prof. D. M. Dhamdhere (CSE) (for Operating Systems, and his high ethical standards), Krishna Kaliappan (Chem), Amit Agrawal (MECH), Devang Khakhar (ChE) and Milind Sohoni (optimization) have all been inspirations. The younger folk like Suyash Awate (CSE), Manjesh Hanawal (OR), Ajit Rajwade (CSE), S Krishna (CSE) (Verification) and Mythili Vutukuru (CSE) (networks) all have good research potential. Prof. Virendra Singh (EE), has been inspirational in support and advice. In fact, our lab would have long closed down had it not been for Prof. Virendra Singh. Across departments, there are huge talents in IITB, and it has been a pleasure.

10. You were previously with Fujitsu Laboratories of America in Richardson Tx, USA, as a member of research staff (2001–05), and before at Cisco Systems. You have also held positions as a visiting scientist at Massachusetts Institute of Technology. So, how is it different to work abroad and in India? Where did you enjoy the most?

[Ashwin]: It is hard to say, where I enjoyed the most. Every place has its charms and its share of headaches. It has been a good journey so far. There have been challenges, but largely it is enjoyable. The problem with India is that it is so bureaucratic here that it gets difficult to focus on technology at times. Indian admins and decision-makers are used to the concept of a manager and not so much of a concept of a stand-alone, hands-on researcher. But some people like me can never be managers, so we should be let alone to do our research and engineering. Having said that, I also feel that there is huge amount of favoritism, and lobby culture in India. Regional biases are too strong in Indian academia, and this must end for some sort of progress to happen. Government/agencies should look for real experts as opposed to please-all administrators, who will only tow a certain line. There are many people who will start a project, few who actually deliver something tangible as part of a project result. Hence agencies must look for output and outcome, and not at just outlay.

11. How has been your life as a professor at IITB and how has it enhanced you?

[Ashwin]: Oh, when I came here, I was told by someone that perhaps I am not good enough for this place. Though it was a bullying comment, I was naïve and took that comment very seriously and it had a profound impact on me, I stopped voicing any opinion in faculty meetings – that has not changed since. So, in that sense, I still am not sure, if I am worthy to be here because I never asked the person(s) for their opinion thereafter. There is of course scope for improvement -- like my teaching to improve, my research to get better, and products to sell more ?.

12. How does undergraduate student culture here compare with institutions abroad where you have been associated? You might have some beautiful memories and incidents during your student times. Any particular you wish to share?

[Ashwin]: The talent of the UG student community here is second to none. My concern is that they do not apply it in research and entrepreneurship as much as I would have liked them to. The country needs sustainable, applicable research. Taking smart ideas and technology to the community is important. New companies should be based on products and new technologies, not on innovative business processes. UGs must not look at IITB as just a degree granting institution, but one from where they can build technologies to make unicorns. UGs should also focus more on products and technologies than short term jobs in finance and other service sectors.

13. Would you like to share your first encounter with research?

[Ashwin]: Way back as a grad student 22 years ago, I was exposed to optical networking as part of a grad course. I am still passionate about optical networking and networks in general. It was just a class assignment, and I went on to read papers in the area of optical networks. Today, all of your data goes over optical fibers. Fibers are entering offices, homes. Today of course, people are launching 600 Gbps/800 Gbps per channel, and 48 such channels in a fiber. Since then, I have changed my direction a bit, towards the systems that make the communication smarter – routers and Software Defined Networks though I still continue to work in Optical Networks.

14. As a professor, what qualities do you expect from a student who is willing to pursue research in your field? 

[Ashwin]: Dedication, hard-work, collaborative skills. Of course, being in CS/EE, one must be good in math, algorithms and coding. Software is eating the world.