Depositphotos
Researcher at the Department of Computer Science at Dartmouth College in the United States Siddhartha Jayanti asked how the computer networks at great distances in deep space.
Will such networks be effective if installed on spacecraft or satellites? As noted by Jayanti, at the heart of distributed computing, the key is efficient communication between different computers on the same network, transmitting data to each other. To mathematically test the performance of a distributed network algorithm, experts usually actually stop its operation at different points in time in order to examine its state and behavior at a given moment and understand how it has evolved over time.
«But what if we now consider a scenario in which these machines are deployed throughout the solar system in spacecraft traveling at high speeds and subjected to unusual gravitational effects? What if different machines are affected by different gravitational fields?», — Jayanti asks.
Currently, scientists are already thinking about creating an idea interplanetary Internet network, that would efficiently and quickly transmit information in space. According to Jayanti, the difference from the operation of such a network on Earth will be that the space network will be affected by the effects of General Relativity described by Albert Einstein. Effects that distort our perception of space and time must be taken into account by specialists who will design distributed networks for operation in space.
In particular, depending on the position of the Earth and Mars, light can take from 3 to 22 minutes to travel the distance between the planets. Because of this, the operation of a distributed space network will be asynchronous. Messages and events between networked computers that are remote from each other will be difficult to coordinate with the different time dimensions they are in.
As noted by Jayanti, for distributed networks, «relativity of simultaneity» — that is, the agreement of two observers of two events occurring in different places that they are happening simultaneously. And this will depend on how fast each of them is moving relative to the location where the events are taking place. This one the effect becomes noticeable only when the analyzed velocities are a significant fraction of the speed of light. This means that observers and computers aboard spacecraft traveling at different speeds will disagree about the order of events.
«According to Einstein, there is no universal freezing of time. So, when our methods of reasoning about distributed systems depend on stopping at different points in time, how do we design algorithms that behave correctly, and how do we verify that they do?», — Jayanti asks.
In a recent scientific paper, published in the Proceedings of the ACM, Siddhartha Jayanti establishes a connection between the properties of classical, relativistic, and computational execution of distributed algorithms. In his article, Jayanti considers many algorithms that have been proven to be correct in classical distributed systems and transfers them to scenarios in which observers monitor their execution from different reference frames, while the machines of the distributed system move at relativistic speeds.
«The surprising result of the paper is that if the algorithm you use is classically correct, then every observer will agree that it is correct in relativistic terms. At the same time, observers may disagree about why the algorithm is correct», — the scientist emphasizes.
Our arguments Jayanti is building on Causality. According to this principle, the cause must precede the effect, even in conditions where the principles of relativity apply. The central idea is to correctly formulate the notion of causality in distributed computing, which is independent of physics, and to establish a connection between this purely mathematical concept and relativistic causality, which is a real physical concept, and to unite them.
The results of the study are published in the journal Proceedings of the ACM
Source: TechXplore