Keysight Technologies Unveils 1.6T Ethernet Test Platforms for AI Data Center Infrastructure

Keysight Technologies Introduces 1.6T Ethernet Test Platforms

Keysight Technologies announced two 1.6T Ethernet test platforms designed to validate next-generation AI data center and hyperscale networking infrastructure1

1

2

2

. The Switch and Edge Router Tester (SERT) 1600GE and AresONE Plus 1600GE platforms were unveiled at the Open Compute Project (OCP) Global Summit in San Jose, running from October 12-151

1

. These platforms address the growing need for rigorous performance validation as AI infrastructure scales to unprecedented levels of complexity and bandwidth demands.

AresONE Plus 1600GE Sets Performance Benchmarks

The AresONE Plus 1600GE establishes a new benchmark by delivering eight 1.6T Ethernet ports in a single 2U appliance while providing stateful, realistic AI workload generation with mix-of-experts emulation1

1

. This platform packs eight OSFP1600 ports into a single 2U chassis, delivering 12.8T of line-rate traffic at 64 bytes2

2

. The platform supports 1600GE PAM4 signaling over 224G SerDes with backward compatibility to 112G, plus flexible port breakout configurations of 1x1600GE, 2x800GE, 4x400GE, or 8x200GE1

1

. AresONE Plus 1600GE can dynamically switch between an eight-port mode for data center switch qualification and a four-port mode for AI benchmarking on the same chassis, helping teams use one appliance across multiple test scenarios4

4

.

SERT 1600GE Targets Edge Router Validation and Switch Fabric Testing

SERT 1600GE targets data center switch fabric and edge router validation, available in an eight-port 2U configuration for switch fabric scale testing2

2

. Both SERT 1600GE and AresONE Plus 1600GE are capable of being time synchronized using Precision Time Protocol (PTP) for larger topologies, offering flexibility to scale test environments as network configurations grow1

1

. The combined platforms provide a consistent test foundation for device qualification, routing, and AI fabric validation, enabling performance, interoperability, and control-plane evaluation as networks scale1

1

.

Protocol Support and AI Workload Emulation Capabilities

Both platforms support Border Gateway Protocol (BGP), Equal-Cost Multipath (ECMP), Virtual Extensible LAN (VXLAN), and SONiC validation across routing, bridging, and access use cases2

2

. AresONE Plus 1600GE integrates with Keysight AI Data Center Builder software for mix-of-experts emulation, RoCEv2, Priority Flow Control (PFC), Explicit Congestion Notification (ECN), and Ultra Ethernet Consortium (UEC) support1

1

. These capabilities let customers exercise both network protocols and AI traffic behavior in the same validation environment to identify performance and congestion issues before deployment4

4

. Both platforms support DAC, ACC, AEC, AOC, and coherent optics in both IHS and RHS form factors2

2

.

NVIDIA Collaboration and Industry Interoperability Demonstrations

Through collaboration with NVIDIA, Keysight AresONE Plus 1600GE provides the high-performance validation platform needed to stress NVIDIA Spectrum-6 Ethernet Switch capabilities with 1.6T/800GE traffic throughput, 224G SerDes, routing-scale validation, and realistic AI workload emulation4

4

. Gilad Shainer, Senior Vice President of Networking at NVIDIA, stated: "AI factories require an Ethernet fabric designed for accelerated computing, and NVIDIA Spectrum-X provides the performance, predictability and control needed to keep large-scale AI workloads running efficiently. Keysight's validation capabilities help engineering teams emulate realistic AI traffic and test Spectrum-X-based fabrics before deployment, reducing time to market"1

1

. Keysight Technologies is also demonstrating multi-vendor interoperability of Link Layer Retry technology with Broadcom and Alibaba Cloud at the OCP Global Summit3

3

. The demonstrations showcase Link Layer Retry operation across 200G and 100G SerDes implementations, along with Credit-Based Flow Control at 100G SerDes, designed for Ethernet-based AI scale-up networks3

3

. One demonstration involves Link Layer Retry running over 200G SerDes-based 1.6T speed between Keysight's INPT-1600GE and a Broadcom Tomahawk 6-based switch, while another shows Link Layer Retry and Credit-Based Flow Control running over 100G SerDes-based 400GE speed with an Alibaba switch built on the Broadcom Tomahawk Ultra ASIC3

3

.

Why This Development Matters for AI Infrastructure

Ram Periakaruppan, Vice President and General Manager of Network Test & Security Solutions at Keysight Technologies, emphasized the shift in hyperscale infrastructure requirements: "The AI era breaks the rules for hyperscale infrastructure. We have moved beyond fault-tolerant commodity scaling to highly synchronized, specialized AI clusters where every packet, link, node, and nanosecond counts. As engineering teams and operators race to upgrade AI infrastructure to 1.6T Ethernet, they need to emulate the broader AI system throughout the development process -- from early design and device qualification to system-level validation and AI rack deployment"1

1

. These 1.6T Ethernet test platforms enable AI infrastructure validation by providing the tools to test at scale while using the same test foundation across the entire design-to-deployment continuum. The ability to identify performance and congestion testing issues before production deployment reduces risk for organizations investing heavily in AI data center buildouts. As AI systems scale to larger accelerator domains, the validation of hyperscale networking infrastructure becomes critical to ensuring that AI workloads run efficiently without network bottlenecks that could undermine the massive computational investments being made across the industry.