Mellanox Nvidia Infiniband Quantum HDR 200Gb/s QM8700 & QM 8790 Smart Edge Switch

Mellanox has introduced a cutting-edge switch, known for its remarkable intelligence and innovative technology called Co-Design Scalable Hierarchical Aggregation and Reduction Protocol (SHARP). This breakthrough technology empowers in-network computing capabilities. Both the QM 8700 and QM 8790 switches offer an impressive non-blocking bandwidth of up to 16 terabits per second, ensuring lightning-fast communication with a minimal port-to-port latency of under 130 nanoseconds.

KEY FEATURES

• 40 HDR 200Gb/s port and 80 HDR100 100 Gb/s Ports in 1 RU
• 16Tb/s aggregate switch throughput
• Up to 15.8 billion messages per second
• Enables offloading of collective operations from the CPU to the switch network
• 1+1 redundant & hot-swappable power & N+1 redundant & hot-swappable fans
• Makes the fabric capable of self-healing autonomy
• Collective offloads (Mellanox SHARP technology)
• Quick easy setup and management

DATA CENTRES WITH HDR 200G INFINIBAND: As servers become more powerful and are equipped with next-generation processors, the demand for data bandwidth in High-Performance Computing (HPC) environments and Enterprise Data Centers (EDC) is rapidly increasing. This escalating need for bandwidth is driven by faster servers, sophisticated applications that require complex computations, and high-performance storage solutions. To address this growing demand, Mellanox is introducing its next-generation HDR InfiniBand technology, featuring high-speed, intelligent switches that will deliver the maximum available bandwidth.

TECHNICAL FEATURES

Infiniband Smart HDR Edge Switches

QM8700 and QM8790 switches are built using Mellanox Quantum InfiniBand switch technology, delivering an impressive capacity of up to forty 200Gb/s ports with full bi-directional bandwidth per port. These standalone switches are exceptionally well-suited for various applications, including top-of-rack leaf connectivity and the construction of clusters spanning from small to extremely large sizes. Both models represent the pinnacle of intelligent network switches globally, designed to facilitate in-network computing through the revolutionary Co-Design Scalable Hierarchical Aggregation and Reduction Protocol (SHARP) technology. The Co-Design architecture empowers the utilization of all active data center devices to enhance communication frameworks using embedded hardware, resulting in dramatic improvements in application performance.

These switches excel in promoting efficient computing, boasting a range of features such as static routing, adaptive routing, congestion control, and enhanced VL mapping to accommodate contemporary topologies like SlimFly, Dragonfly+, and 6DT. These features ensure optimal utilization of fabric bandwidth by mitigating congestion hotspots. Moreover, these switches are designed with a strong emphasis on energy efficiency, featuring best-in-class power consumption. Additionally, power consumption is further minimized when ports are partially utilized.

The primary distinction between the two models lies in their management modes. The QM8700 offers a control interface for out-of-band management, whereas the QM8790 relies on the NVIDIA Unified Fabric Manager (UFM) platform for management purposes.

The self-healing capabilities of NVIDIA Mellanox InfiniBand prevent link failures and increase network performance by 5,000x when compared with software-based solutions—improving system performance, scalability, and network utilization.

COMMUNICATION ACCELERATION

Collective communication acceleration pertains to communication patterns involving all members of a group of communication endpoints. This type of communication is frequently employed in High-Performance Computing (HPC) protocols like MPI and SHMEM.

The Mellanox Quantum switch enhances the efficiency of specific collective operations by handling data as it moves through the network. This eliminates the necessity to transmit data repeatedly between endpoints. Additionally, it facilitates the aggregation of substantial data vectors at optimal transmission rates, enabling crucial MPI large vector reduction operations, particularly vital in machine learning applications.

HDR100

A QM8700 or QM8790 HDR 100 module coupled with a Mellanox ConnectX-6 adapter card provides HDR100 support. Two pairs of two lanes per port allow these switches to support up to 80 ports of 100G, making them the densest TOR switches available. For racks with more than 40 servers per rack, this is a perfect solution. It can also be used for small to medium deployments that need to scale to a 3-level fat tree in order to reduce power consumption, latency, and space consumption.

MANAGEMENT

ComEx Broadwell X86 CPU in the QM8700 includes onboard subnet managers, allowing an out-of-the-box fabric setup for up to 2K nodes. Through CLI, WebUI, SNMP, or JSON interfaces, it offers full chassis management using MLNX-OS.

Mellanox's QM8700 and QM8790 integrate with its Unified Fabric Manager (UFM) software to manage scale-out, and InfiniBand computing environments efficiently and monitor cluster health. Using UFM, the fabric is up and running at maximum capacity at all times.

High-performance clusters and data centers can be deployed with these QM8700 series switches. This range of InfiniBand switches offers superior performance, low power and scale, reducing capital and operating expenses, and delivering the highest return on investment when evaluating price versus performance or energy usage.

POWERING FEATURES

ORDER INFORMATION

Manufacturers are responsible for warranties, which are based on purchase date and validity. We make sure the products we sell are delivered on time and in their original condition at Grabnpay. Whether it's physical damage or operational issues, our team will help you find a solution. We help you install, configure, and manage your devices. We'll help you file a warranty claim and guide you every step of the way.