Spirent Communications, a provider of test and assurance solutions, has introduced test solution capable of emulating Artificial Intelligence (AI) workloads over Ethernet.
This innovation addresses a need in the networking domain, where Ethernet serves as the linchpin of Cloud infrastructure, necessitating the ability to accurately emulate AI traffic workloads and assess their impact on AI data center networks and interconnects.
“AI represents a pivotal focus area for Spirent as applications like ChatGPT, Lensa, Copilot, and others reshape the communications landscape,” Eric Updyke, CEO of Spirent Communications, said.
“Hyperscale cloud providers are reallocating investments from conventional data center front-end operations to revamped back-end infrastructures tailored for managing the surge in AI applications and workloads.”
The evolution of AI-driven environments necessitates infrastructure that is distinct from traditional data centers, catering specifically to the demands of AI. Testing the Ethernet fabric of these novel AI environments has presented formidable challenges. AI training imposes unique requirements, including heightened workload volumes, acute sensitivity to latency, congestion, and job completion times (JCT).
“As data centers reimagine their architectures for AI, testing assumes paramount importance. One of Spirent’s top priorities has been developing a specialized test solution capable of seamlessly emulating realistic xPU workloads and AI traffic patterns. Our new solution empowers engineers to evaluate their Ethernet fabric without the need for costly investments in new xPU server-equipped labs,” Eric Updyke said.
Spirent’s AI test solution, operating on the A1 400G platform, represents the inaugural hardware capable of emulating high-density 400G xPU workloads tailored for AI environments. This innovation allows customers to assess their Ethernet fabrics within existing test environments, obviating the need for constructing expensive new xPU server-equipped labs.
Leveraging the RoCEv2 protocol, the Spirent platform offers intuitive configuration, facilitating the generation of AI workloads using real servers and ensuring repeatable, consistent results to streamline the testing of AI use cases. Moreover, it functions as a versatile platform capable of concurrently testing AI and routing/switching use cases, enhancing its utility and value proposition.
