Harnessing Best-Effort Computing to Enable Dynamic Artificial Life Simulations at Scale

Emerging ML/AI hardware accelerators, like the 850,000 processor Cerebras Wafer-Scale Engine (WSE), hold great promise to scale up the capabilities of evolutionary computation. However, challenges remain in maintaining visibility into underlying evolutionary processes while efficiently utilizing these platforms’ large processor counts. Here, we focus on the problem of extracting phylogenetic information from digital evolution on the WSE platform. We present a tracking-enabled asynchronous island-based genetic algorithm (GA) framework for WSE hardware. Emulated and on-hardware GA benchmarks with a simple tracking-enabled agent model clock upwards of 1 million generations a minute for population sizes reaching 16 million. This pace enables quadrillions of evaluations a day. We validate phylogenetic reconstructions from these trials and demonstrate their suitability for inference of underlying evolutionary conditions. In particular, we demonstrate extraction of clear phylometric signals that differentiate wafer-scale runs with adaptive dynamics enabled versus disabled. Together, these benchmark and validation trials reflect strong potential for highly scalable evolutionary computation that is both efficient and observable. Kernel code implementing the island-model GA supports drop-in customization to support any fixed-length genome content and fitness criteria, allowing it to be leveraged to advance research interests across the community.

Matthew Andres Moreno, Connor Yang, Emily Dolson, and Luis Zaman. 2024. Trackable Island-model Genetic Algorithms at Wafer Scale. In Proceedings of the Companion Conference on Genetic and Evolutionary Computation (GECCO ‘24 Companion). Association for Computing Machinery, New York, NY, USA. https://doi.org/10.1145/3638530.3664090

• LaTeX source via GitHub
• primary publication

Continuing improvements in computing hardware are poised to transform capabilities for in silico modeling of cross-scale phenomena underlying major open questions in evolutionary biology and artificial life, such as transitions in individuality, eco-evolutionary dynamics, and rare evolutionary events. Emerging ML/AI-oriented hardware accelerators, like the 850,000 processor Cerebras Wafer Scale Engine (WSE), hold particular promise. However, practical challenges remain in conducting informative evolution experiments that efficiently utilize these platforms’ large processor counts. Here, we focus on the problem of extracting phylogenetic information from agent-based evolution on the WSE platform. This goal drove significant refinements to decentralized in silico phylogenetic tracking, reported here. These improvements yield order-of-magnitude performance improvements. We also present an asynchronous island-based genetic algorithm (GA) framework for WSE hardware. Emulated and on-hardware GA benchmarks with a simple tracking-enabled agent model clock upwards of 1 million generations a minute for population sizes reaching 16 million agents. We validate phylogenetic reconstructions from these trials and demonstrate their suitability for inference of underlying evolutionary conditions. In particular, we demonstrate extraction, from wafer-scale simulation, of clear phylometric signals that differentiate runs with adaptive dynamics enabled versus disabled. Together, these benchmark and validation trials reflect strong potential for highly scalable agent-based evolution simulation that is both efficient and observable. Developed capabilities will bring entirely new classes of previously intractable research questions within reach, benefiting further explorations within the evolutionary biology and artificial life communities across a variety of emerging high-performance computing platforms.

Moreno, M. A., Yang, C., Dolson, E., & Zaman, L. (2024). Trackable Agent-based Evolution Models at Wafer Scale. In The 2024 Conference on Artificial Life. MIT Press. https://doi.org/10.1162/isal_a_00830

• manuscript source via GitHub
• kernel software via GitHub
• data via Open Science Framework ❋
• conference slides via Google Slides

Here, we test the performance and scalability of fully-asynchronous, best-effort communication on existing, commercially-available HPC hardware.

A first set of experiments tested whether best-effort communication strategies can benefit performance compared to the traditional perfect communication model. At high CPU counts, best-effort communication improved both the number of computational steps executed per unit time and the solution quality achieved within a fixed-duration run window.

Under the best-effort model, characterizing the distribution of quality of service across processing components and over time is critical to understanding the actual computation being performed. Additionally, a complete picture of scalability under the best-effort model requires analysis of how such quality of service fares at scale. To answer these questions, we designed and measured a suite of quality of service metrics: simulation update period, message latency, message delivery failure rate, and message delivery coagulation. Under a lower communication-intensivity benchmark parameterization, we found that median values for all quality of service metrics were stable when scaling from 64 to 256 process. Under maximal communication intensivity, we found only minor – and, in most cases, nil – degradation in median quality of service.

In an additional set of experiments, we tested the effect of an apparently faulty compute node on performance and quality of service. Despite extreme quality of service degradation among that node and its clique, median performance and quality of service remained stable.

Moreno, M. A., & Ofria, C. (2022). Best-Effort Communication Improves Performance and Scales Robustly on Conventional Hardware. arXiv preprint arXiv:2211.10897.

• data via Open Science Framework ❋
• source code via GitHub
• manuscript source via GitHub

Developing software to effectively take advantage of growth in parallel and distributed processing capacity poses significant challenges. Traditional programming techniques allow a user to assume that execution, message passing, and memory are always kept synchronized. However, maintaining this consistency becomes increasingly costly at scale. One proposed strategy is “best-effort computing”, which relaxes synchronization and hardware reliability requirements, accepting nondeterminism in exchange for efficiency. Although many programming languages and frameworks aim to facilitate software development for high performance applications, existing tools do not directly provide a prepackaged best-effort interface. The Conduit C++ Library aims to provide such an interface for convenient implementation of software that uses best-effort inter-thread and inter-process communication. Here, we describe the motivation, objectives, design, and implementation of the library. Benchmarks on a communication-intensive graph coloring problem and a compute-intensive digital evolution simulation show that Conduit’s best-effort model can improve scaling efficiency and solution quality, particularly in a distributed, multi-node context.

Matthew Andres Moreno, Santiago {Rodriguez Papa}, and Charles Ofria. 2021. Conduit: a C++ library for best-effort high performance computing. In Proceedings of the Genetic and Evolutionary Computation Conference Companion (GECCO ‘21). Association for Computing Machinery, New York, NY, USA, 1795–1800. https://doi.org/10.1145/3449726.3463205

• data via Open Science Framework ❋
• interactive data notebooks via MyBinder ❋
• graph coloring benchmark source code via GitHub
• digital evolution benchmark source code via GitHub
• slides via Google Slides
• LaTeX source via GitHub

Developing software to effectively take advantage of growth in parallel and distributed processing capacity poses significant challenges. Best-effort computing models, which relax synchronization requirements, have been proposed as a strategy to overcome challenges harness high performance computing at extreme scale. Although many programming languages and frameworks aim to facilitate software development for high performance applications, existing prevalent tools do not expose an explicit best-effort interface. The Conduit C++ Library aims to provide a convenient interface for best-effort inter-thread and inter-process communication. Here, we describe the motivation, objectives, design, and implementation of the library.

Matthew Andres Moreno, Santiago Rodriguez Papa and Charles Ofria. 2021. Conduit: A C++ Library for Best-Effort High Performance Computing. MSPDS 2020: The 6th International Workshop on Modeling and Simulation of and by Parallel and Distributed Systems.

• poster
• LaTeX source via GitHub

Matthew Andres Moreno, Santiago Rodriguez Papa, and Charles Ofria. 2021. Conduit: a C++ library for best-effort high performance computing. In Proceedings of the Genetic and Evolutionary Computation Conference Companion (GECCO ‘21). Association for Computing Machinery, New York, NY, USA, 1795–1800. https://doi.org/10.1145/3449726.3463205

• documentation via ReadTheDocs 📖
• source archive via Zenodo z