Skip to product information
1 of 1

Arista

Arista DCS-DS-7720R4-128PE-F | 7700R4 Distributed Spine Switch, 128x 800G OSFP

SKU:DCS-DS-7720R4-128PE-F

Regular price POA
POA
POA
Sale Sold out
Taxes included. Shipping calculated at checkout.

Description

The Arista DCS-DS-7720R4-128PE-F is a high-performance distributed spine switch designed for ultra-scalable AI and machine learning clusters. Featuring 128 x 800GbE OSFP fabric ports and built on advanced 5nm Jericho3-AI silicon, this system operates as part of the 7700R4 Distributed Etherlink Switch architecture to deliver lossless, non-blocking performance for demanding accelerated computing workloads. The bundle includes Supervisor-EL module with 8-core CPU, 64GB RAM, 512GB NVMe storage, and front-to-rear airflow configuration with 4 AC power cords.

Features

Distributed Etherlink Switch (DES) architecture for logical single-hop forwarding
- Broadcom Jericho3-AI silicon with AI-optimized packet pipeline
- Massively parallel distributed end-to-end scheduling
- Cell-based traffic spraying fabric for congestion-free operation
- Virtual Output Queuing (VOQ) eliminates head-of-line blocking
- Distributed credit scheduling ensures fairness across all flows
- 100% internally lossless forwarding with deep buffering
- Support for RDMA over Converged Ethernet (RoCE)
- DCQCN, PFC, ECN, and Cluster Load Balancing support
- Ultra Ethernet Consortium (UEC) standards compatible
- Advanced telemetry with flow-level visibility via CloudVision
- Arista EOS network operating system
- Hot-swappable power supplies and fan modules
- NEBS compliance ready with optional grounding kit
- 5nm silicon technology for maximum power efficiency
- Linear scalability from hundreds to tens of thousands of ports
- Support for passive DAC cables up to 4m reducing power consumption
- Linear-drive Pluggable Optics support for lower power footprint

Warranty

All products sold by XS Network Tech include a 12-month warranty on both new and used items. Our in-house technical team thoroughly tests used hardware prior to sale to ensure enterprise-grade reliability.

All technical data should be verified on the manufacturer data sheets.

View full details

specs-tabs

Collapsible content

Technical Specifications

FAQs

Technical Specifications

128 x 800GbE OSFP fabric ports for distributed spine connectivity
- Based on Broadcom Jericho3-AI 5nm silicon architecture
- Supervisor-EL module: 8-core/16-thread 2.3 GHz CPU
- Memory: 64GB RAM, 512GB NVMe storage
- Virtual Output Queuing (VOQ) architecture eliminates head-of-line blocking
- Cell-based traffic spraying fabric with distributed credit scheduling
- 100% internally lossless and fair forwarding
- Front-to-rear airflow (-F suffix)
- Includes 4 x AC power supplies with SAF-D C20 cords (3m)
- 2RU form factor with 4-post mounting kit
- Support for passive copper cables up to 4m and Linear-drive Pluggable Optics
- Part of distributed system supporting over 4,600 x 800GbE hosts in 2-tier configuration
- Ultra Ethernet Consortium (UEC) ready
- Arista EOS operating system with CloudVision management

FAQs

Q: What is the Arista 7700R4 Distributed Etherlink Switch architecture?
A: The 7700R4 DES is an ultra-scalable distributed system that operates as a single logical switch across multiple physical devices. It combines dedicated leaf and spine switches using a 2-tier topology to create a massively parallel fabric supporting over 4,600 x 800GbE hosts in a single configuration, with logical single-hop forwarding for predictable, lossless performance in AI/ML clusters.

Q: What workloads is the DCS-DS-7720R4-128PE-F designed for?
A: This spine switch is optimized for large-scale AI training and inference workloads, machine learning clusters, and HPC environments requiring RDMA over Converged Ethernet. It handles synchronized high-bandwidth collective communication operations like All-Reduce and All-to-All across thousands of GPU/XPU accelerators with deterministic, congestion-free performance.

Q: How does the 7720R4 spine switch integrate with leaf switches?
A: The 7720R4 spine connects exclusively to 7700R4 series leaf switches via its 128 x 800G OSFP fabric ports. Spine switches do not connect to each other or external networks. The distributed architecture uses automated hardware-accelerated scheduling and deterministic traffic spraying across all leaf-spine interfaces to achieve 100% path utilization without manual tuning.

Q: What advanced features support AI cluster performance?
A: The switch implements Virtual Output Queues (VOQ) to eliminate head-of-line blocking, cell-based traffic spraying for optimal load balancing, distributed credit scheduling for fairness, and deep buffering to handle incast patterns. It supports DCQCN, PFC, ECN, and Cluster Load Balancing protocols essential for RDMA workloads, with forward compatibility for Ultra Ethernet standards.

Q: What is included in the -F bundle configuration?
A: The DCS-DS-7720R4-128PE-F bundle includes the 7720R4 chassis with 128 x 800G OSFP ports, Supervisor-EL module with 8c/16t CPU, 64GB RAM and 512GB NVMe storage, front-to-rear airflow fans, 4 x AC power supplies with 4 x SAF-D C20 power cords (3m length), and 4-post mounting hardware. The -F suffix indicates fixed front-to-rear airflow direction.

Recently Viewed

  • Request a Quote

    Looking for competitive pricing? Submit a request, and our team will provide a tailored quote that fits your needs.

  • Contact Us Directly

    Have a question or need immediate assistance? Call us for expert advice and real-time support.

    Call us Now  
  • Contact Us Directly

    Have a question or need immediate assistance? Call us for expert advice and real-time support.

    Contact us