Bright Computing Plans to Deliver Exascale Cluster Management in 2020

Bright Computing, provider of platform-independent cluster management software, has plans to deliver exascale cluster management software in the first half of 2020.  

Exascale high-performance computing systems will be capable of executing at least one exaflop, or a quintillion (1018) floating point operations per second.

In 2015, the Exascale Computing Project was created as part of the National Strategic Computing Initiative in the U.S. with the goal of building an exascale computer by 2021. Most recently the U.S. Department of Energy announced its plans to procure multiple exascale supercomputers at a total cost of up to $1.8 billion.

Earlier this year, the European Commission announced the EuroHPC Joint Undertaking designed to deploy exascale computing infrastructure in Europe. The commission’s contribution of €486 million represents nearly half of the projected €1 Billion needed to deliver exascale computing by 2022-2023.

Others including Japan and India are funding significant efforts on research and development towards the design and building of a state-of-the-art exascale supercomputer.

Significant advances in computing performance—from giga to tera to peta—have been achieved through a combination of increased chip performance and increased parallelism. However, the step from peta- to exa- is expected to derive almost entirely from parallelism. Exascale systems will contain vast numbers of processors and nodes compared to today’s petascale systems.  

Bright’s exascale-capable version of Bright Cluster Manager is designed to support 100,000+ nodes. The company began delivering enhancements towards exascale in 2016 with features such as dedicated provisioning nodes, and a new monitoring subsystem designed for extreme scale. Current development work includes dedicated monitoring nodes, hierarchical object rendering in the Bright UI, optimized API communication patterns and exascale simulation testing.

Among the top 10 technical challenges facing the development of viable exascale systems are energy efficiency, scalable system software and scientific productivity. Bright Cluster Manager software is designed to help solve problems in all of these areas.

“Our software’s ability to monitor a cluster’s health and quickly identify problems ensures clusters remain online and functioning effectively,” says Martijn de Vries, CTO at Bright Computing. “Its ability to scale to over 100,000 nodes will enable our partners to develop exascale supercomputers that take advantage of Bright Cluster Manager’s powerful management capability and ease of use at scale never before seen.”

By making Bright Cluster Manager exascale-ready well before the first systems are scheduled for delivery, Bright’s partners and customers will be able to take advantage of Bright’s advanced management and monitoring capabilities to ease their system development and testing efforts, accelerate deliveries and create systems with less administrative burden.

“Bright Cluster Manager automates most of the procedures involved in building and running clusters,” says Robert Stober, director of Product Management at Bright Computing. “The ease of management that our software provides increases the productivity of computational scientists by providing them with the power tools they need to effortlessly keep their computing systems running smoothly.”  

Bright will be providing additional details on its exascale program at the Supercomputing conference in Dallas, TX, November 12-15 (booth #3025).

Sources: Press materials received from the company.