Keysight and AMD Break Barriers to Redefine Benchmarking for Cloud and Edge Infrastructure Performance

• Solution tests mobile and 5G private network performance with millions of concurrent devices establishing simultaneous voice, video, and web sessions in real-world scenarios

• Supports a wide variety of industry use cases like smart manufacturing and smart energy grids that must connect millions of low-latency, high-bandwidth devices

• New methodology uses realistic traffic generation to uncover a CPU’s true power and scalability while observing bandwidth requirements, enabling better network provisioning

KEYSIGHT Technologies, Inc. (NYSE: KEYS) leveraged 4th generation AMD EPYC™ CPUs from Advanced Micro Devices, Inc. (NASDAQ: AMD) to develop an industry first benchmarking methodology that redefines the landscape of system performance evaluation for high-speed digital designers, network equipment manufacturers, and data center operators. 

The deployment of mobile and 5G private networks is expected to grow rapidly across many industries as part of their automation and digital transformation strategy. This diverse set of industry use cases requires a versatile compute architecture that can simultaneously support millions of low-latency, high-bandwidth devices, and a multifaceted workload profile. Given this transformation, comprehensive infrastructure benchmarking is critical to ensure optimal performance across a spectrum of industry use cases.

To provide more realistic performance benchmarks, Keysight in collaboration with AMD developed an integrated test case methodology that takes into account the processing power of central processing units (CPUs). Using realistic traffic emulation, this novel approach enables data center operators to extract more performance from a chosen CPU and to characterize the performance across multiple vectors while adhering to bandwidth requirements. The methodology was developed by deploying Keysight's software tools onto AMD EPYC™ processors to conduct staged test cases with realistic traffic at scale.

Highlights of the test case methodology include:

• Infrastructure scalability – In settings such as smart grids and smart manufacturing, smooth and reliable operation requires handling millions of concurrent devices. Traditional benchmarks frequently overlook the CPU's capacity and its capability to handle a considerably higher concurrent load as the network's bandwidth constraints obscure this ability. This new benchmarking approach is designed to uncover the true CPU power and scalability in these environments.
  

• Real-life authenticity – Mobile and 5G private networks support specialized fields like vision-based smart manufacturing, extended reality (XR) viewership, and retail workflows. Testing infrastructure for its capability to handle realistic application, voice, and video traffic is crucial. The new benchmark emphasizes replicating bulk activities with millions of devices establishing new sessions and including highly realistic scenarios like subscribers initiating voice, video, and web calls. This insight into real-world capabilities empowers data center operators to provision their infrastructures accurately.

  
  

Raghu Nambiar, Corporate Vice President, Data Center Ecosystems and Solutions, AMD, said: "5G mobile network operators worldwide are facing an unprecedented demand for bandwidth. This consistent growth of data traffic volume requires scalable and effective benchmarking solutions to accurately measure user experience. Our collaboration with Keysight demonstrates the ability of AMD EPYC CPUs to deliver impressive parallel processing performance to handle these high volumes of data traffic."

Ram Periakaruppan, Vice President and General Manager, Network Test & Security Solutions, Keysight said: "Single dimension benchmarking methodologies like throughput are insufficient for data center operators to properly design and deploy their compute infrastructure. New performance benchmarking is needed in the modern mobile and 5G private networks. When used, it highlights the prowess of the AMD 'Zen 4' chiplet architecture to meet complex demands in those networks."