When I first started using computers, the mainframe was the only game in town. Computing was centralized at my university with everyone, faculty, students, and administration sharing the same machine. It would often get bogged down with too many requests, meaning hours of delay before getting your output. Decentralized computing was a welcome advance as computing could be done on our own personal desktop machines. As the internet has expanded we are in the position where the volume of data being created is once again is threatening to overwhelming centralized approaches. To centralize or decentralize computing is a vital question. A recent article in Science Magazine by Rebecca Napolitano, Wesley Reinhart and Juan Pablo Gevaudan explains recent decentralized advances to remedy this problem.
To Centralize or Decentralize Computing
The early days computing was entirely centralized. Computers were too large and expensive to exist outside of data centers. If you wanted to computerize your data analysis, you had to submit your job via punch card deck and wait, sometimes for hours, for the job to run. At my university, the computer would be periodically upgraded, but it was a struggle for the computer center to stay ahead of the increasing demand for computing. By the late 80s personal computers were on the scene, which allowed for decentralized computing. You could run your analysis on your own desktop device, so no more waiting for output from the computer center. Those early desktop computers were relatively expensive, but today quite capable devices such as the Raspberry Pi, can be purchased for just a few dollars.
Over time there have been forces moving in both directions—to both decentralize computing via personal devices and to centralize by tying all those devices together. For example,“Big Data” projects require so much computing power, that they have to be conducted across hundreds and even thousands of individual machines. Single mainframes have been replaced with server farms that allow many users to complete tasks from a centralized array of linked computers. For some applications, such as playing an online video game, there is no choice. You must log into the centralized server to play the game. For other applications there can be choices about whether to work centralized on a remote system or decentralized on your own device.
New Decentralized Technologies to Reduce Centralized Overload
Napolitano and her colleagues explain how smart cities are incorporating advanced information technology to improve the quality of life for residents. This means monitoring information about everything from strain on a building to citizen input about policy decisions. New capabilities make it possible to incorporate sensors into building materials to monitor when maintenance needs to be done. For example, a building might have sensors that can detect settling cracks in a wall or foundation. Advanced materials might also be self-healing. If such sensors were streaming data to a centralized computer, it would soon be overwhelmed, making such an approach impractical.
Two concepts the authors note is “fog computing” and “mist computing”. Fog computing is semi-centralized and occurs when a series of devices sends data, not to a central server, but to an intermediate computer that analyzes and synthesized the data. Thus, each building might have a central computer to handle its own data. Mist computing is totally decentralized with each sensor having its own microprocessor to analyze its data. These strategies reduce the load on centralized servers as well as the bandwidth needed to handle massive amounts of data.
As computing has developed, the demand for the processing of more and more data has skyrocketed. This means we need to have the ability to centralize when information needs to be in one place and decentralize when it does not. As data demands continue to increase we will see advances on both the centralized and decentralized fronts.
Paul Spector is a professor in the School of Information Systems and Management, University of South Florida. He teaches executives and does research on the human side of organizations. He blogs on this site about technology and mental health and here about the human side of working.