Changes coming

http://www.businessweek.com/magazine/content/07_52/b4064048925836_page_5.htm

This can lead to huge changes in the way consumers, and ultimately businesses, do computing. And I don't mean the web browsing kind of computing. On the personal computing front google et al have already put forward their vision: everyone would have a box with little computational power and a broadband connection. Then all the actual work a typical user does will be done on a remote server with the box serving as the monitor interface. This is the kind of dumb-terminal mainframe approach of fifteen years ago, but probably will be cheaper than a PC with the added advantage that the user wouldn't have to upgrade his PC every two years.

Industrial designers happen to use huge computing power today. VLSI design, automotive design, aircraft engine design and several others are examples. All design cycles are, well, cyclic. You conceive a product, you do some back of the envelope (or equivalent) calculation, you define the architecture (usually in a divide and conquer approach, by breaking the system into smaller subsystems, and then into individual blocks). Till this point not a lot of computation are needed. But beyond this point serious computational power is usually needed, so much so that it becomes a bottleneck. After this point individual blocks have to be designed to their individual specs and verified. Usually there are iterations to check manufacturability and reliability. Then the blocks are put together in sub-systems and verified again. At this point some block specs may change and design iterations done again. Also exhaustive verification for a sub-system may be too computationally expensive and in many cases is not possible because computational resources will be too expensive to aggregate.
In the end sub-systems are put together to form systems, and these are not even verified with the actual sub-systems present. At this stage models of sub-systems' behaviour are used to simulate the systems. That is itself is extremely taxing on the computers.

Several, if not all, of the verification problems can be solved if more powerful computers are available to the designers. However several of the computationally toughest problems are not frequent enough for the design companies to invest in hardware and memory, thanks to the fact that design is cyclic. If the design companies can rent some computational power in such times that will be extremely beneficial to them.

In addition cost of infrastructure (hardware and software) is one of the biggest barriers to entry in the design field for entrepreneurs. A rental service will tackle such costs, and this means entrepreneurial activity all over the world can take off in design space.

I will eagerly wait to see what this move harbingers.

Frankly Internet was a more or less useless place till about 2000 (save the email), then wikipedia, amazon etc. made it useful. Fads like Facebook existed ten years ago, and went away with time. The real usefulness of the information revolution lay in connecting information held by people, not people themselves who were already connected by phone and email.

The present great change set to happen is death of communications as we know them once VoIP through technologies such as WiMAX takes over. That will be the next great thing happening on the Internet. It'll become The Communication Channel, by having all of the world's communications passing through it.

The foreseeable future is elaborated to some extent in the Businessweek story at the top. Internet will become a utility, a computational and communication utility, for everyone. And that will be the next revolution, a true IT revolution.