Marketplace for computations

Distributed computing affordable.

In my senior year as a Computer Science undergrad,  I asked the guest speaker about “why it was not yet possible to access sizeable computing power and resources on the internet at very low costs. This given that there were more idle computers on the internet that used computers.” Professor V Rajaraman considered the founding father of computer science education in India was the chief guest, and he gave me an amazing answer – about how if one were to lift a large object say 100 times one’s weight, it would need lock-step coordination. He gave the example of ants and the swarm of ants being able to – in lockstep, carry an object many times their weight because they were able to coordinate perfectly.

 

Computation Power for hire

Later, as generations of computers were built, there was the whole mainframe revolution, then a distributed cluster revolution, and, more recently in the past decade a cloud revolution that started offering such clustered-coordinated-computations as a marketplace. Pay-as-you use models, became possible, though the price of computation seemed exorbitant for students, and firms with very low budgets.  obviously, as with any large infrastructure based marketplace a few top players dominate.

Golem

Golem, a distributed app, running on the ethereum network is one such solution. It allows a pay-per-use model for renting as many nodes as needed to run applications. Both the service providers i.e. individuals who are willing to let other people use idle computing time on their systems, and, customers of the service are bound by an ethereum smart contract. Though Golem is in its very early stages with a Beta being released on the Ethereum Main-net, this distributed app is extremely promising if it succeeds in accomplishing what it has set out for. The kinds of distributed apps, that need high-end computing ranges from developing 3-dimensional graphical applications for games, to, enabling deep learning through massively parallel analytical tools. Today, such infrastructure is accessible only at research universities, large firms or governments. Renting such infrastructure say 10 nodes on the cloud would set one back by a few 100 dollars per day of usage, based on the usage. Golem’s fee-based revenue model will hopefully cost much less for computational usage, and, will possibly democratize and make computation pervasive.

The baby steps were made this past week, but, over the next few years, the possibilities of this are immense. If price parity and network scaling for smart contracts do not bottle neck the growth of this app, a few years down the line, this could truly challenge the dominance of large cloud based infrastructure players, who will either have to switch to a similar model or alter the focus of their businesses. This application is decentralization at work.