Atanu Dey On India's Development

India and Utility Computing

Stand-alone computing a la PCs delivering “services” is fine for those who can afford that luxury, but is definitely a show-stopper for those who have very little disposable income and yet can make use of those services that PCs deliver. I remind myself repeatedly that people do not want a PC — what they actually want are the services that a PC delivers. As long as we focus on the fact that it is services — and not the hardware nor the software — that matter to people, we will not end up putting the cart before the horse. So if a firm were to deliver those set of services at an affordable price, it is immaterial to the consumer whether the consumer (of those services) uses a PC or some other device.

We know that low costs translate into low prices. How does one reduce costs? If there are economies of scale in production, then centralizing the production is the obvious answer. A pertinent example is that of electric power production. Each consumer could have a generator at home. But it is much cheaper if a centralized facility generated the power at a much lower cost per unit due to scale economies and distributed the power to the consumers on an as-needed basis. There is one tradeoff here which involves the cost of distribution.

How large should a centralized power station be depends on the distribution costs and on the scale economies. Distribution costs depends on the density of consumers and on the distribution technology. That tradeoff therefore implies that in a large city (very high consumer density) a large centralized power plant is required. But for a large number of consumers spread over a very large area, the distribution costs may overwhelm the production economies of scale, such as would be the case in rural areas. So smaller plants with smaller distribution networks will be more effective in rural areas. In the limiting case, each consumer will have to have their own generators.

Computing as a utility depends on scale economies and on the distribution costs as well. And the distribution costs depend on the technology of distribution and on the consumer density. Consumer density is critical because the distribution costs are significant today because broadband is expensive.

Consumers who are most able to use computing services but cannot afford to go the route of having PCs number in the tens of millions in India. For instance, there are small and medium businesses (SMBs), and educational institutions such as schools and colleges. Because they are “consumer clusters”, computing as a utility is very appropriate for them.

Here is a thumbnail description of a utility computing platform. The central server forms the core where you have a very wide range of software applications, plus a massive collection of rich content (audio, video, text, and graphics) and storage. The server is accessed over a local area network (LAN) using access devices that are inexpensive and easy to manage. The access devices are sometimes refered to as “thin clients” — a device that hangs off the LAN and is connected to a display, keyboard, and a mouse. The TCs do not have local storage. Centralizing the production of computing services on the server has numerous advantages, most notably that of taking the management of the hardware/software resources required for the user services out of the hands of the users.

There is hardly anyone who has ever used a connected PC and not been frustrated by problems such as viruses, spam, spyware, the need to frequently upgrade hardware and software, and so on. Users have come to expect that these problems are a necessary part of using computers. It need not be so. It is a bit of a mystery why people put up with the bother and inconvenience of using computers. Imagine if you had to open up the hood every few days and tinker around the car’s innards trying to fix some problem or the other. You would quickly dump that sort of car for something that works without you getting your hands dirty.

If using computing services were to become more like the telecommunications services model, then more people would be able to use them. You sign up for the service, and you pay every month for your usage. You let the firm supplying you the service to fix things if things break.

Utility computing is the future. For more on the topic, see Rajesh Jain’s Emergic Blog.

You may ask, how is utility computing relevant to India’s development. I will tell you. The future of India depends on education. India will not develop unless we can educate the hundreds of millions that need it. Resources are limited and one of the best ways of leveraging limited resources is to use information and communications technology (ICT) tools. Schools and colleges which cannot afford the PC-centric solution need utility computing services.

My day job for the last few months has been to figure out a solution to the challenge of delivering education. I think that I have a set of answers that I would go into later.

  • http://aatman.blogspot.com Anant

    Indeed…the next big thing would be some sort of a utility computing wherein people have access to the resources they do not own.

    Grid Computing, which is still in the research phase, is coming up on similar lines wherein people would be able to share not just the processing cycles or hard disk space but also centralized devices that would be able to cater to the needs of hundreds of clients, thus removing redundancy in the sense that every individual client need not install its own centralized device.

  • http://uspeed.blogspot.com uspeed

    Utility computing is fine and its an elegant solution, but an economic case for such thin clients has to be made. For instance, what shall be the costs of connecting the clients to the central servers ? What about the cost of providing data security to communications ? Reliability ? Scalability ? Who shall bear the cost of the servers ? Besides, the cost of “production of services” doesnt go down as more and more services are produced. It goes up as more and more exotic hardware is used to support greater production of services. The marginal utility added per $ spent on harware decreases rapidly once one moves beyond the optimal point.

    Anyhow, I think its a fairly involved excercise to work out the economics and technology challenges involved in the solution. Would love to read more about it if theres anything online.

  • http://uspeed.blogspot.com uspeed

    I was assuming I wud check ur comments section again for ur reply. Didnt leave my email as I dont want it to be harvested by bots :-)

    Its j s u d e e p [at] y a h o o . c o . i n

    (no spaces, no underscores, nothing).

    Ill go through the IBM journal and use ur comment box to gas after Ive done so :-P