Theoretical foundations of software systems
Research Article
Buran: A parallel dataflow computation model based on token interaction
Arkady Valentinovich Klimov
| National Research Center „Kurchatov Institute“, Moscow, Russia | |
|
|
Abstract.
An abstract computation model is presented, which is a generalization of a family of architectures and models based on token interactions.
Tokens are viewed as abstract entities. Interactions between them generate computational quanta, which can generate new tokens.
The model is generally nondeterministic. However, an individual computation can be deterministic.
This article provides a mathematical description of the model, defining the concepts of computation and deterministic computation.
The fundamental decidability of the determinism of a given computation is proven,
but the question of how to recognize it more efficiently remains open.
Being abstract, the model can have various concretizations.
One of them is the project of the unique multi-core architecture of the BURAN parallel dataflow computing system (PDCS),
proposed by V.S. Burtsev in the late 1990s, as well as some of its extensions.
The name of the abstract model itself was borrowed from this project.
Work on it is currently ongoing at the Kurchatov Institute National Research Center.
The closest known analogue of the model are Petri nets and their extensions, such as Colored Petri Nets (CPN).
They can also be described as a concretization of the abstract model. (In Russian).
Keywords: abstract computation model, parallel computation model, determinism, nondeterminism, token, interaction, computational graph, PDCS BURAN, Petri nets
MSC-2020
68Q09; 68T01, 68W10Acknowledgments: The work was carried out within the state assignment of NRC „Kurchatov institute“
For citation: Arkady V. Klimov. Buran: A parallel dataflow computation model based on token interaction. Program Systems: Theory and Applications, 2026, 17:4, pp. 3–32. (In Russ.). https://psta.psiras.ru/2026/4_3-32.
Full text of article (PDF): https://psta.psiras.ru/read/psta2026_4_3-32.pdf.
The article was submitted 28.05.2026; approved after reviewing 10.09.2026; accepted for publication 26.09.2026; published online 10.10.2026.