- VAST Data collaborates with Cisco to integrate with Nexus 9000 Ethernet switches, enhancing networking for AI datacenters.
- Cisco’s Silicon One ASIC chips power its fastest network switches, delivering high throughput and advanced capabilities.
- Nexus HyperFabric AI clusters, leveraging Nexus switches, assist in building AI datacenters with Nvidia-accelerated computing.
- Cisco certifies Nexus switches with VAST Data’s storage solutions, enabling seamless integration and monitoring.
- Collaboration simplifies AI implementation for enterprises, easing the transition from proof of concept to full-scale production.
Main AI News:
In a strategic partnership, VAST Data pioneers integration with Cisco’s Nexus 9000 Ethernet switches, fortifying the networking giant’s Nexus HyperFabric for Ethernet AI stacks linked to Nvidia GPU farms.
Cisco, a frontrunner in networking solutions, has engineered groundbreaking Silicon One ASIC chips for its swiftest network switches and routers. The June 2023 G200 product, boasting 512 x 100GE Ethernet ports on a single device, delivers an impressive 51.2 Tbps throughput. It encompasses advanced congestion management, packet-spraying techniques, and robust link failover capabilities. The Silicon One hardware is available as standalone chips or chassis devoid of Cisco software. Cisco has successfully implemented its Ethernet AI Fabric, powered by Silicon One, at three hyperscale clientele.
The Nexus HyperFabric AI clusters, leveraging Nexus switches, are meticulously crafted to aid enterprises in erecting AI datacenters employing Nvidia-accelerated computing prowess, including Tensor Core GPUs, BlueField-3 DPUs, SuperNICs, and AI Enterprise software, all orchestrated through Cisco networking infrastructure and the VAST Data Platform. Notably, the VAST system seamlessly integrates BlueField-3 DPUs. Cisco has certified Cisco Nexus Ethernet-based switches with VAST Data’s storage solutions, furnishing validated designs. Moreover, Cisco customers can seamlessly monitor and correlate storage performance and latency through VAST’s APIs, seamlessly integrating both network and storage telemetry within the Nexus HyperFabric ecosystem.
Renen Hallak, CEO and co-founder of VAST Data, asserts, “This marks the year of enterprise AI adoption. Historically, enterprises grappled with the intricacies of integrating novel technologies due to the complexities of harmonizing them with existing systems and processes. The collaboration with Cisco and Nvidia simplifies AI implementation for enterprises, ushering them from proof of concept to full-scale production seamlessly.”
Jonathan Davidson, Cisco EVP and GM for networking, echoes this sentiment, stating, “With an ecosystem bolstered by VAST Data and Nvidia, Cisco empowers enterprise clientele to navigate their most formidable networking, data, and security challenges, fostering AI infrastructures of any magnitude with unwavering confidence.”
The Cisco Nexus HyperFabric boasts a myriad of features:
- Streamlined IT operations facilitated by Cisco’s cloud management capabilities across all workflow phases.
- High-performance spine and leaf switches from the Cisco Nexus 9000 series, offering 400G and 800G Ethernet fabric performance.
- The Cisco Optics family, featuring QSFP-DD modules, affording customers unparalleled choice and superlative densities.
- NVIDIA AI Enterprise software, streamlining the development and deployment of production-grade generative AI workloads.
- NVIDIA NIM inference microservices, expediting the deployment of foundation models while upholding stringent data security protocols, available alongside NVIDIA AI Enterprise.
- NVIDIA Tensor Core GPUs, commencing with the NVIDIA H200 NVL, engineered to turbocharge generative AI workloads with unparalleled performance and memory capabilities.
- BlueField-3 data processing unit DPU processor and BlueField-3 SuperNIC from NVIDIA, accelerating AI compute networking, data access, and security workloads.
- An enterprise-ready AI reference design built on NVIDIA MGX, a modular and adaptable server architecture.
- The VAST Data Platform, encompassing unified storage, database, and a data-driven function engine tailored for AI applications.
Furthermore, the BlueField-3 DPUs are adept at executing security services such as the Cisco Hypershield, facilitating an AI-native, hyperdistributed security paradigm wherein security is intricately interwoven with workload exigencies.
VAST and Cisco assert their commitment to simplifying network management and operations across diverse infrastructure endpoints. The private cloud (on-premises) managed Cisco Nexus Dashboard offers congestion management with precision flow control algorithms and real-time telemetry for enhanced visibility.
A select cohort of customers can now access the VAST Data Platform in conjunction with Cisco and Nvidia on a beta basis, with widespread availability slated for calendar Q4 of 2024.
Conclusion:
The collaboration between VAST Data and Cisco, bolstered by Nvidia, signifies a significant step forward in enterprise AI adoption. This partnership streamlines AI implementation for enterprises, addressing integration complexities and facilitating smoother transitions to production-grade AI infrastructures. The integration of VAST Data’s storage solutions with Cisco’s networking infrastructure underscores a commitment to simplifying network management and operations, ultimately enhancing the efficiency and effectiveness of AI datacenters. This development is likely to reshape the market landscape, making AI adoption more accessible and manageable for a broader range of enterprises.
