BioPaul Borrill is an engineer/entrepreneur building a software infrastructure company in Silicon Valley based on algorithms whose assumptions about causality go beyond Newtonian and Minkowski spacetime. Our product is a new protocol based on FPGAs that addresses fundamental problems in distributed systems. We plan to use Mathematica to design and verify our protocols, which are (effectively) multiway rewriting rules that compile to C or C++ expressions and are executed on FPGAs.
Problem: Maintaining liveness and synchronizing processes in network channels is a perennial challenge when packets can be dropped, reordered, duplicated or delayed. Conventional switched networks require protocol stacks and applications to use timeouts and retries to maintain liveness. This makes exactly-once semantics impossible and precipitates retry storms that lead to unbounded tail latency and transaction failure.
Our architecture is a “network automata.” We employ individual direct channels with the Stop and Wait (SAW) protocol to maintain flow control, Colored Petri Nets for conserved quantities and the Spekkens Toy Model (STM) to manage epistricted consistency for distributed metadata.
Our software infrastructure is a “stacked tree” network model for distributed computing. We would like to build models of causal structures that can be securely executed at extreme low latency in the FPGA “substructure” of a datacenter.