The future of high-performance data centers is here

network-fabric

In today’s data centers, traditional networks are now the single biggest bottleneck, throttling applications, driving up costs and creating unsustainable environmental impacts. With layers of expensive switches and inefficient design, spine and leaf architectures inherently congest, starve resources and dramatically reduce performance at scale.

The Rockport scale fabric addresses this bottleneck head-on by directly interconnecting nodes using pre-wired supercomputing topologies. High performance without the headaches, the fabric is fast to deploy, easy to scale, transparent to manage, and moves data from source to destination faster than traditional networks, making the entire data center more efficient.

Looking for the highest performance per watt per dollar in the industry? Welcome to the fabric of the future.

 

 

Video transcript

With the industrialization of high-performance computing, more and more organizations are adopting increasingly sophisticated infrastructure to solve complex problems and deliver results faster.

Compute and storage technologies have evolved to address these challenges. However, relying on decades-old spine and leaf architectures, networking has lagged behind.

With layers of expensive switches and inefficient design, traditional high-performance networks inherently congest, slowing applications, underutilizing both compute and storage resources, and dramatically impacting performance at scale.

In modern data centers, traditional networks are now considered the single biggest issue throttling applications, driving up costs and environmental impacts.

Rockport’s new approach to high-performance networking accelerates applications by moving data from source to destination faster and more efficiently than ever before. We started by eliminating the switches. We used advanced software to embed the network inside each endpoint. The result is a fully distributed network fabric designed for the data-intensive AI/ML era. This next-generation fabric directly interconnects the nodes using pre-wired supercomputing topologies to address congestion and latency head-on.

The network is self-discovering, self-optimizing, and self-healing. So, it’s fast to deploy and easier to maintain and scale without user intervention. It continually optimizes the best path through the network to avoid congestion and latency. And breaks the data into smaller packets called FLITs so the network can handle low latency and high bandwidth traffic at the same time.

Every source/destination pair has 12 potential routes that can be optimized, both by job and by individual packet. An Autonomous Network Manager provides users with health status, performance metrics, error counters, and more to see what’s happening in the network and troubleshoot issues.

Rockport has redefined networking for both compute and data-intensive environments with a new architecture that enables better performance and with the right economics. The Rockport fabric delivers accelerated and predictable performance, has superior compute and storage utilization, and offers a sophisticated yet simple user experience. High performance without the headaches. The fabric is fast to deploy, easier to scale, transparent to manage, and moves data from source to destination faster than traditional networks. Making the entire data center more efficient.

And a Rockport fabric does it all with the highest performance per watt, per dollar in the industry. Welcome to the Fabric of the Future.

Here’s something similar we think you’ll like.