Keysight AI Data Center Builder is designed to provide an efficient, cost-effective, comprehensive, and precise solution for testing and validation of AI data centers. Its development focuses on meeting market demands for AI workload simulation, network performance verification, multi-scenario adaptability, and product quality enhancement. It is committed to expanding its applications in the field of test and measurement through innovative technologies, supporting the verification and evaluation of new converged computing and networking technologies, thereby establishing a significant technical position in the AI data center construction wave and driving the industry toward more efficient and reliable network infrastructure.
• 800G/112G SerDes Rocev2 Load Simulation: Leading one generation ahead of network cards, this simulation capability helps validate and evaluate Rocev2 switches that have already entered the 800G capability level.
• Rank Allocation: It allows assigning specific rank IDs to test ports for participation in collective communications. This enables verification across various scenarios, such as track optimization and multi-tenancy affinity, without altering the physical connections.
• Statistical Analysis Based on Queue Pairs: Statistical analysis based on queue pairs holds significant guiding value in many scenarios. For instance, it can directly verify the fairness of network bandwidth in the context of DCQCN flow control, thereby improving cache allocation and watermark settings to optimize balance.
• Automatic DCQCN Tuning: Based on automated scripts, this feature conducts automatic testing of DCQCN parameters on the client side, searching for optimal combinations to provide reference for DCQCN tuning on the NIC.
• Simultaneous Measurement of Module Quality and AI Load Performance on a Single Platform: It enables rapid measurement of common retransmissions and failures in the current network caused by module oscillation, ACK timeout, and other issues, quickly identifying segmentation problems.
• Support for Multi-Tenant Simulation: It allows simultaneous simulation of parallel measurements for different tasks from multiple tenants, enabling pressure testing on AI infrastructure in a configurable and controllable manner.