OTTAWA, Ontario, October 26, 2021 – Rockport Networks today announced the commercial availability of its new switchless network architecture that delivers the industry-leading performance and scalability needed for high-performance computing workloads, including HPC, l ‘IA and ML. Addressing long-standing data center bottlenecks that hold back innovation, the Rockport switchless network offers an all-new design, using breakthrough data routing software and techniques to overcome congestion, delivering Predictable performance improvements of more than 3 times that of centralized, highly switched networks.
The Rockport Switchless Network distributes the network switching function to end devices, where those devices (nodes) become the network. By eliminating switch layers, the Rockport switchless network also significantly frees up rack space for better compute and storage utilization, as well as savings in power, cooling, and associated administrative costs. . Compute and storage resources no longer run out of data, and researchers have more predictability around workload runtime.
“Rockport was founded on the fact that switching and networking in general is extremely complicated. Over the years, this complexity has forced organizations to compromise when it comes to performance at scale, so we decided to keep it simple, ”said Doug Carwardine, CEO and co-founder of Rockport Networks. “We’ve made it our mission to get data from source to destination faster than other technologies. Removing the switch was crucial to achieving significant performance benefits in an environmentally and commercially sustainable manner.
Redesigning network switches creates an opportunity to take advantage of direct interconnect topologies that provide a mesh of connectivity in which each endpoint in the network can efficiently forward traffic to all other endpoints. The Rockport Switchless Network is a distributed, highly reliable, high-performance interconnect providing pre-wired supercomputer topologies through a standard plug-and-play Ethernet interface.
The Rockport Network Operating System (rNOS) is the software at the heart of the Rockport Switchless Network and runs on the network card, fully off the server’s compute cores and operating system. RNOS allows the network to self-discover, self-configure and self-repair. As a navigation application for data, this patented approach constantly selects and optimizes the best path through the network to minimize congestion and latency, while breaking down packets into smaller chunks (FLIT) to ensure that messages from high priority are not blocked by large or bulk messages. data transfers. Designed to easily integrate into existing and emerging data centers, the Rockport Switchless Network installs in a fraction of the time required to wire traditional switched networks. As an embeddable architecture, it will work in any form factor. Today, software is deployed and managed using three main components:
- The Rockport NC1225: The NC1225 is the industry’s first true network card, replacing standard network interface cards (NICs) in servers and storage enclosures, as well as the switch layers that form traditional network structures. It adaptively aggregates the bandwidth of multiple parallel network paths, taking advantage of 300 Gbps of available network capacity.
- The Rockport SHFL: SHFL eliminates months of cabling with a single optical device, this new invention of passive cabling offers end users a simple way to implement sophisticated supercomputing network architectures in a fraction of the time of switched networks. The Rockport SHFL is available in several versions for easy scaling of networks.
- The Rockport Autonomous Network Manager (ANM): The ANM offers a holistic view and an in-depth view of the active network. It continuously monitors all aspects of the network and stores historical data for 60 days, with the most recent seven days being collected with high fidelity in the timeline.
Rockport’s technology offers significant benefits in terms of total cost of ownership, durability and safety, including:
- Eliminates the need for redundant devices for reliability
- Dematerializes and decarbonises the data center, thus reducing the carbon footprint
- Requires 34% to 69% less power consumption compared to switching technologies
- Eliminate up to 72% of wiring needs
- Reduces rack space and weight in the network by up to 77%
- Protects against security vulnerabilities resulting from continuous re-optimization of data paths
“When the root of the problem is architecture, building a better switch just didn’t make sense,” said Matt Williams, CTO, Rockport Networks. “With sophisticated algorithms and other specially designed software advancements, we’ve solved congestion issues so our customers no longer need to spend bandwidth on their network issues. We focused on real world performance requirements to set a new standard for what the market should expect from fabrics of the future.
After an extensive beta program for cloud service providers and elite government and academic research labs, the Rockport Switchless Network is being rolled out by customers including the University of Texas Advanced Computing Center (TACC). . The company also works with industry organizations, including Ohio State University (OSU) to contribute to high-performance networking standards.
TACC Establishes New Center of Excellence with Rockport Networks
In today’s news, Rockport Networks is announcing a collaboration with TACC to create a Center of Excellence in Austin, Texas. TACC is home to Frontera, the fastest supercomputer on a college campus, and the 10e most powerful supercomputer in the world. TACC installed 396 nodes on Frontera running production workloads on Rockport’s switchless network, including quantum computing, pandemic-related life science research as well as workloads focused on rapid responses to emergencies such as hurricanes, earthquakes, tornadoes, floods and other large-scale disasters.
“TACC is very pleased to be a Rockport Center of Excellence. We run a variety of advanced IT workloads that rely on high bandwidth, low latency communication to maintain performance at scale. We are excited to be working with innovative new technologies such as Rockport’s switchless network design, ”said Dr. Dan Stanzione, Director of TACC and Associate Vice President for Research at UT-Austin. “Our team is seeing promising first results in terms of congestion and latency control. We were impressed with the ease of installation and management. We look forward to continuing to test new and larger workloads and further expanding the Rockport switchless network in our data center. “
“The Durham team continues to push the boundaries when it comes to discovering next-generation HPC networking technologies,” said Alastair Basden, DiRAC / Durham University, technical director of the COSMA HPC cluster. “Based on a 6D toroid, we found the Rockport Switchless Network to be remarkably easy to configure and install. We have looked at codes that rely on point-to-point communications between all nodes with varying packet sizes where, typically, congestion can reduce performance on traditional networks. We were able to achieve consistent low latency under load and are eager to see the impact this will have on larger scale cosmology simulations. “
“There is significant evidence to suggest that switchless architectures have the ability to dramatically improve application performance, which has traditionally been very expensive,” said Earl C. Joseph, CEO of Hyperion Research. “Making these advancements more economically accessible should greatly benefit the global research community and hopefully improve the expectations of what we can expect from the network in terms of research return and lead time. getting results. “
“Our mission is to provide the advanced computing community with standard libraries such as MVAPICH2 that support the best possible performance available in the market. We make it a top priority to keep our libraries fresh with innovative approaches, like the new switchless architecture from Rockport Networks, ”said DK Panda, professor and eminent scholar of computer science at Ohio State University, and head of research in network-based computing. Group. We look forward to our continued partnership with Rockport to set new standards for our next releases. “
Availability and additional information
The Rockport Switchless Network is available immediately.
For more information, including full product details, please visit: https://rockportnetworks.com.
Rockport Networks launch activities
Following the launch of the Rockport Switchless Network, the company will participate in a number of industry events, including:
Virtual roundtable with HPCwire – November 10: Rockport joins Dell, TACC, OSU and Hyperion Research for a virtual discussion: “Quieting Noisy Neighbors: Keys to Addressing Congestion in Multi-workload Environments”. Register here.
Rockport Networks at SC21 – November 14-19: Rockport is a sponsor of the SC21 Virtual Exhibition. Rockport experts will be available to answer questions in the virtual exhibit hall and Rockport Technical Director Matthew Williams will present: “Advanced congestion control to resolve network performance bottlenecks using of a new generation interconnection ”, with a live question-and-answer session. Learn more at www.rockportnetworks.com/sc21.
About Rockport Networks
Rockport Networks next generation of high performance networks unlocks the entire data center to deliver more results, faster, with better economy and environmental sustainability. Inspired by the world’s fastest supercomputers, the Rockport switchless network replaces centralized switch architectures with a high-performance distributed direct interconnect that automatically discovers, configures, and repairs, and is simple and transparent to use. By virtually eliminating congestion and latency, data center workloads can run much faster, enabling organizations to improve ROI and make critical decisions faster. Learn more at www.rockportnetworks.com.
Source: Rockport Networks