LinkedIn 
Contact Us 
X Rockport Ethernet Fabric Tech Primer
Learn more about the Rockport Ethernet Fabric’s unique distributed architecture.
Copyright © Cerio 2023
Cerio is a trade name of Rockport Networks Inc.
Built for modern data centers
The Rockport Ethernet Fabric goes beyond the limits of traditional network architecture to accelerate AI and HPC innovation. Rockport’s distributed, direct-interconnect design combines with advanced congestion controls to deliver faster results, higher resource utilization and sustainability across the data center.
28%
faster workload completion (on average)
3.5X
reduced latency under load
Optimize your clusters
Limit congestion under load, achieve better workload completion times and strand fewer resources.
Save time and money
Deploy a network in a fraction of the time, while reducing power, cooling, rack space and cabling.
Scale as you go
Start with what you need and reuse nodes for a completely modular, flexible and cost-effective network fabric.
See inside your network
View real-time, end-to-end traffic flows and troubleshoot network performance.
Distributed by design. Optimized for performance.
The Rockport Ethernet Fabric delivers the best performance per watt per dollar in the industry.
By distributing the network switching function into each device endpoint, the nodes become the network, for advanced congestion control, linear, elastic scaling and built-in redundancy.
Key features
NC1225 Network Card
- 300Gbps fabric per node
- 12 fiberoptic connections to the network (25Gbps each)
- 8 best paths aggregated and continually optimized
- Patented distributed FLIT switching for superior performance
- Field-Upgradeable FPGA, so the network always has the latest features
- Standard Ethernet host interface (verbs and sockets)
Network Operating System (rNOS)
- Makes the network self-discovering, self-configuring, and self-healing
- Selects and continually optimizes the best path to minimize congestion and latency
- Breaks packets down into smaller pieces (FLITs) to ensure high-priority messages are not blocked by large messages or bulk data transfers, further reducing latency
SHFL Optical Interconnect
- A passive optical interconnect that delivers plug-and-play, pre-wired supercomputing topologies in a box
- Stunningly simple cabling solution for rapidly deploying large scale networks
- 6 months of cabling pre-wired
- Requires zero power or cooling
- No specialized training required
Autonomous Network Manager (ANM)
- Turn up your network automatically
- Monitor the performance of individual nodes
- Track end-to-end traffic flows
- Manage system-wide settings
- Optimize advanced features like ultra-high priority
- See it in action: watch the demo
Our Resources
Get the tech primer
Dispelling Bandwidth Myths in High Performance Networks
Find out why bandwidth is no longer the currency for performance.
The Consequences of Congestion
It’s time to resolve the root cause of congestion.