Computing as a Commodity



Computing as a Commodity: Distributed Computing over the Global Internet

Imagine an Internet that automatically and securely carries out complex computational tasks for geographically dispersed users, serves their rich personalized data requests, and provides seamless group communication.  How do we model this new paradigm to enable algorithms that are secure, efficient, robust, and energy aware?


Today’s Internet serves not only as a communication network but also as a computing tool. Increasingly clients are separated from their data and processing resources, while we use Internet-based applications for remote services like ordering merchandise and filing tax returns. It is natural to envision computing power as a future commodity. Research in theory of computing can lead these new trends (a) by mathematically modelling the heterogeneity of connected computers and the diversity of computing tasks, as well as their security and economics, (b) by creating new metrics that capture energy limitations and communication bottlenecks, and (c) by developing solutions for seamlessly and transparently distributing dynamic workload across the network. Such research will enable collaborations between theoreticians and more applied computer scientists.


The modern computing environment has changed significantly over the past decades. Starting from a collection of computers loosely connected over a local file system, our data is now distributed all over the world. Moreover, we increasingly use Internet-distributed applications to process this data. It is natural to believe that soon much of our computation will be performed inside the network, and that computing power will be a commodity, like water and electricity. We will be able to purchase over the network not only faster machines and better connections, but also raw computing power.

Tomorrow’s network will be more heterogeneous, consisting not only of server farms, but also personal computing devices like PCs, PDAs, and even small low-power sensors. This ubiquity and heterogeneity of computational devices gives rise to many algorithmic challenges: How do we store the data so that it is readily accessible when we need it? How do we ensure the security and privacy of the data? How should the diverse tasks originated network-wide be scheduled in the network subject to the various computing and communication constraints? How do we price the services to encourage collaborative behavior among the users and the service providers? How do we ensure reliability and provide protection across multiple layers and heterogeneous computing environments?

Tomorrow’s network will also need to be energy aware, expanding the computing power under severe energy constraints. How do we exploit the trade-off among speed, energy and performance? What is a good metric, beyond addition to space and running time, that captures the energy issue? How do we measure communication that increasingly dominates other costs within processors and among computers and memories?

New models that allow an explicit role for aforementioned new issues will reinvigorate numerous applied fields of computing, placing them on a firm and mutually collaborative foundation. It will advance our understanding and future development of data-driven computation, computing-as-commodity, and farming of data over the global Internet.

Contributors and Credits

Anupam Gupta, Rajmohan Rajaraman, Udi Wieder, David Wise, Lisa Zhang

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: