Arrcus Enables Actapio with Next-Gen Networking Fabric to Build-out AI Clusters for Large Language Models

High-performance, lossless network fabric from Arrcus with advanced capabilities, on UfiSpace switches, to maximize performance from GPUs deployed for Training Models

BARCELONA, Spain--(BUSINESS WIRE)--Arrcus, the hyperscale networking software company and a leader in core, edge, and multi-cloud routing and switching infrastructure, has provided Actapio, a subsidiary of LY Corporation, cutting-edge networking infrastructure to build-out AI clusters. Leveraging Arrcus' ACE-AI solution and open networking hardware from UfiSpace, built on Broadcom Tomahawk 5 switching silicon, Actapio is now equipped to train and deploy large language models (LLMs) deployed on Nvidia GPUs with unprecedented efficiency, enhancing user experiences for millions of LY Corporation users.

“We are excited to support Actapio in driving innovation in AI infrastructure with our ACE-AI solution,” said Arrcus Chairman and CEO, Shekar Ayyar. “This deployment demonstrates the power of Arrcus’ cutting-edge networking solutions to address the complex needs of modern AI workloads, enabling enterprises to unlock new levels of performance and efficiency.”

To support its LLM capabilities, Actapio required a robust GPU cluster capable of processing massive datasets over extended training cycles. Transitioning from Hadoop-based server applications with 25-50 Gbps interfaces to Nvidia GPU servers with 100-400 Gbps interconnects presented challenges, including network bottlenecks, energy optimization, and minimizing operational costs. Arrcus' ACE-AI networking solution addressed these demands by providing a unified, modern approach to optimizing GPU clusters and distributed compute resources, seamlessly integrating infrastructure from edge to core to multi-cloud.

Key highlights of the ACE-AI solution include:

• Built on Broadcom’s Tomahawk 5 silicon for lossless Ethernet with RoCEv2, PFC, ECN, and Dynamic Load Balancing.
• Support for Non-blocking 51.2 Tb/s on a single chip.
• Easily scale to 512 x 400G ports (and more) for GPU server connectivity with a Leaf-Spine CLOS fabric.
• Open networking hardware from UfiSpace platform (S9321-64E).
• Ensures resiliency, availability, and visibility with intelligent congestion control and reduced latency.
• Open, cloud-native programmable stack used by leading enterprises and service providers worldwide.
• Open interfaces and standards-based APIs with observability for closed-loop automation.
• Delivers 40%+ TCO savings through operational efficiency and automation.
• Eliminates vendor lock-in, supporting diverse silicon vendors, ODMs, and form factors.

“At Actapio, delivering seamless and personalized experiences to our users is at the heart of what we do. By leveraging Arrcus' ACE-AI solution, we've transformed our infrastructure to meet the rigorous demands of large-scale AI workloads with efficiency and scalability,” said Norifumi Matsuya, CEO and President of Actapio, Inc. “This collaboration allows us to innovate faster and continue enhancing the services we provide to millions of LY Corporation users worldwide.”

By leveraging these advanced features, Actapio has been able to revolutionize its networking infrastructure, creating an environment that is not only high-performing but also streamlined and efficient. This collaboration has delivered measurable results, including:

• 2x Bandwidth, 6x Fewer Switches: Broadcom Tomahawk 5 800G interconnect doubles performance while slashing switch count compared to previous hardware.
• Streamlined Operations: Reduced TCO with minimized overhead, fewer errors, and simpler management.
• Compact and Efficient: Fewer units, optics, and cables cut costs, lower failure rates, and reduce power consumption.

“Broadcom’s Tomahawk 5 silicon provides the robust foundation for Actapio’s next-generation GPU clusters, offering the performance and bandwidth required for advanced AI workloads,” said Ram Velaga, senior vice president and general manager, Core Switching Group, Broadcom. “We are delighted to work with Arrcus and support this innovative partnership with technology that enables enterprises to push the boundaries of AI performance and efficiency.”

“UfiSpace is proud to contribute to this transformative partnership by delivering high-performance, open networking hardware that empowers Actapio's AI clusters,” said Vincent Ho, CEO of UfiSpace. “Our collaboration with Arrcus showcases the potential of open solutions to achieve unparalleled scalability and flexibility in AI networking.”

With the adoption of Arrcus’ ACE-AI solution, Actapio has not only addressed the complexities of modern AI networking but has also set a new standard for efficiency and scalability in AI infrastructure. This partnership underscores Arrcus’ commitment to enabling transformative technology solutions that drive innovation and deliver exceptional value for enterprises.

Additional Resources

• Arrcus Enables Liberty Global’s AI-Powered Telco vision, with Advanced Video Solution for Public Safety, Industrial Automation, and Immersive Entertainment
• Arrcus Newsroom

About Arrcus

Arrcus is a leading provider of networking software solutions that empower businesses to achieve unparalleled scalability, performance, and reliability in their infrastructure. Arrcus is disrupting the industry with disaggregated solutions that deliver innovative, agile, and cost-effective networking, allowing enterprises to break free from traditional, monolithic systems and embrace a more flexible, efficient, and scalable approach to modern networking. The Arrcus team consists of world-class technologists who have an unparalleled record in shipping industry-leading networking products, complemented by industry thought leaders, operating executives, strategic partners, and top-tier VCs. The company is headquartered in San Jose, Calif. For more information, go to www.arrcus.com or follow Arrcus on LinkedIn and Twitter/X.