October 28, 2018

History of Pseudo-Time

  Pseudo-Time, Technology


When NuevoCloud was founded in 2012 our original product was a storage service built on a novel pseudo-time filesystem. Continued development of our filesystem and pseudo-time allowed us to turn it into a distributed system, and eventually build a CDN from it. As far as we know, NuevoCloud is the only pseudo-time system to ever be placed in production.

Pseudo-time was the work of David Patrick Reed at MIT in 1979. Reed was involved in TCP/IP and later went on to design UDP; the two protocols the internet is built on. Despite the success of his later work, pseudo-time was virtually forgotten. I believe pseudo-time was too complex to implement on the hardware of the time, and that limited additional experimentation and exploration of the idea. Reed was too early for his idea to work, and by the time the hardware was ready, the idea was already forgotten.

The building blocks of pseudo-time (possibilities and pseudo-temporal environments) enable new methods of coordinating in a distributed environment. That's why pseudo-time forms the core of our product, and it enables many of the unique capabilities of our CDN.

We'll be talking more about how we use pseudo-time to accelerate network traffic in future articles. For those who want to learn more about pseudo-time, Reed's original paper Naming and Synchronization in a Decentralized Computer System goes over the details of his system.