Arista 7800R4 Series Universal Spine
Designed for the next generation of accelerated computing, large scale virtualized and cloud networks, and service provider backbones, the Arista 7800R4 Series modular switches are the 4th generation of the R-Series universal spine platform. The 7800R4 series combines high density 800G and 400G density with comprehensive functionality for large scale networking as well as dedicated hardware choices for accelerating AI/ML workloads to maximize performance.
Advanced Extensible Operating System (EOS™) features include network monitoring, real-time telemetry, precision timing and EVPN-VXLAN network virtualization. Arista Etherlink® features, combined with Ultra Ethernet compatibility, optimize performance for AI/ML workloads. Together with extensive hardware resources and a wide choice of systems, the 7800 Series provides lasting investment protection.
The 7800R4 Series enables highly scalable and secure network designs with MPLS, Segment Routing, MLAG, ECMP, and EVPN-VXLAN technologies in addition to Arista’s TunnelSec strong encryption for layer 2 and layer 3.
7800R4 Series line cards build on the capability of the 7280R and 7500R Series with support for FlexRoute, Accelerated sFlow and large scale ACLs. FlexRoute provides scalability to support deployment as a routing platform with Internet scale routing. Algorithmic ACLs provide flexible pattern matching for access control, policy based forwarding and network telemetry. 7800R4M Series line cards add wire-speed TunnelSec encryption capability while 7800R4K Series offer both encryption and expanded FlexRoute support to over 5M routes. Accelerated sFlow at high density 100G, 400G and 800G provides visibility and programmatic control of traffic steering with no impact on packet forwarding. *
(* availability 1H 2025)
7800R4C line cards are optimized for cluster computing, implementing a simplified, lower latency and energy conserving pipeline while maintaining the core non-blocking, lossless forwarding architecture that enables intensive applications to perform to their maximum.
7800R Architecture Advantages:
- Consistent 7800R4 Series architecture ensures seamless migration and investment protection
- Distributed packet processor scales forwarding capacity linearly
- FlexRoute™ for Internet route table scale with low power draw
- Flexible pipeline for programmable features including Algorithmic ACLs and Inband Network Telemetry
- System wide health monitoring
- High Bandwidth Intelligent and dynamic deep buffers for the most demanding traffic scenarios
- Virtual Output Queues avoid head of line blocking
- Cell based fabric with redundancy for 100% efficiency and consistent low latency and jitter
- Optimized multicast forwarding
- Comprehensive traffic counters with up to 16 classifications for real-time network monitoring and troubleshooting
- QOS Fabric fairness - no hot-spots with mirroring enhancements
Arista EOS
- Single 64 bit binary image
- Fine-grained truly modular network OS
- Scalable control plane for internet peering
- Extensible platform - bash, python, C++ , GO, OpenConfig
- Resilient self-healing software
- Rich Streaming Telemetry solutions
- Over 15 Years of EOS innovations and experience with the R-Series Architecture
Scalable Cloud and Service Provider Designs
- Comprehensive switching and routing feature set with rapid convergence
- Segment Routing and EVPN Services for IX Providers and Cloud WAN
- IPv4 and IPv6 Overlay and Underlay support
- Internet scale IPv4 and IPv6 forwarding tables with over 5M IPv4 routes
- Large scale IPv4 / IPv6, ACLs, and QoS with flexible profiles
- Up to 173Bpps of wire speed forwarding
- CloudVision® provisioning and automation
- VXLAN and 512-way multi-pathing for next generation DC
- LANZ for microburst detection
- Accelerated sFlow for high capacity network security and application monitoring
- Inband Network Telemetry for flow and packet forensics at scale
- CloudVision for Workflow Automation and Orchestration
High Performance Universal Spine
- Non-blocking 460Tbps (920Tbps FDX) of system capacity
- High density 25/50/100/400G/800G Ethernet
- Wirespeed TunnelSec encryption at 25G-800G
- MACsec, IPSsec and VXLANsec
- Scalable to 576 wire speed 800G ports
- Ultra deep buffers up to 32GB per line card
- Class leading latency of 3.5usec
- 28.8Tbps of fabric capacity per line card
- Virtual Output Queues per port and CoS to eliminate head of line blocking
Arista Etherlink for AI
- AI Analyzer powered by AVA
- Advanced collective load balancing and congestion management
- AI workflow integration
- Designed for future UEC deployments
Cloud Grade Routing
- Comprehensive routing: OSPF/ OSPFv3, MP-BGP, IS-IS, PIM, MSDP and MLD
- Transport Signaling : LDP, RSVP-TE, SR/ SR-TE, BGP-LU, BGP-LS
- Services: IPv4 VPNs (RFC 4364), 6PE/6VPE, L2/L3 EVPN , EVPN-VPWS, EVPN (MPLS), EVPN-MPLS Gateway, EVPN-DCI, EVPN-Multicast, VPLS, mLDP, Pseudowires, PTP, E-OAM
- Tunneling: LDP, RSVP, SR, LDP over RSVP, GRE, MPLS over GRE
- Resiliency: ECMP/ UCMP, PIC, RSVP-TE FRR, SRLG, TI-LFA
- QoS: Classification, Flexible rewrite, Ingress Policing, VoQ, Hierarchical-QoS, Shaping
- Support for 400G-ZR, 800G-ZR and the Arista ZR line system for metro DWDM
- Security: MACsec, IPSec, uRPF, ACLs, CoPP, AAA, Dot1X
Arista 7800R4 Series Specifications
The Arista 7800R4 Series of modular switches deliver the industry’s highest performance with 460Tbps (920Tbps FDX) of system throughput to meet the needs of the largest scale accelerated compute clusters, data centers and provider backbones. With a choice of systems the 7800R4 Series provides four, eight, twelve or sixteen line card modules with redundant fabric and supervisor modules, and a range of grid redundant power supplies for AC, -48V DC and HVDC infrastructure.
Designed for continuous operations and with system wide monitoring of hardware and software components, simple serviceability and provisioning to prevent single points of failure the hardware architecture supports high-availability with hot-swap of all components.
Arista 7800R4 Series Architecture
The Arista 7800R4 Series are purpose built multi-chip systems with fully distributed end to end scheduling and coordination - even in the largest 16-slot, 576x 800GbE variant, the platform operates like a single extremely large switching chip that is 100% internally lossless and fair. Four advanced technologies converge to this highly desirable architecture:
- Cell-based fabric - mitigates elephant and mice flows with uniform cells sprayed across all fabric links for 100% efficiency
- Virtual Output Queues (VOQ) - invokes ingress virtual queues to every egress port, eliminating head of line blocking (HOLB)
- Distributed Credit Scheduling - ensures all egress ports are independent eliminating HOLB and noisy neighbor issues
- Deep Buffering - easily handles incast, bursts and speed mismatches without packet loss, keeping TCP running efficiently
This advanced architecture enables the Arista 7800R4 to handle the most demanding workloads with ease. Accelerated compute clusters, data centers and Service Provider networks all benefit from the 7800R family’s ability to handle high bandwidth low entropy traffic, mixed traffic loads, real-time, multicast, and storage traffic while still delivering low latency and non-blocking performance.
7800R4 System Comparisons
The 7800R4 Series offers extremely high density at four levels of capacity, with 16, 12, 8 and 4 slot chassis options available. Supporting a broad range of wirespeed interfaces from 25G to 800G, the 7800R4 series lowers total cost of ownership through efficient design principles. With power consumption as low as 5W per 100Gbps, discrete cooling zones and advanced environmental monitoring to optimize power consumption, the 7800R4 Series delivers maximum reliability, density, and power efficiency.
The 7800R4 Series offers choice of line cards with 36 x 800G interfaces, supporting wire speed breakout to 25G, 50G, 100G, 200G and 400G including support for ZR and ZR+ coherent DWDM transceivers, providing unparalleled flexibility and the ability to seamlessly transition data centers, accelerated compute clusters and provider backbones to the next generation of Ethernet performance.
System Performance | |||||
---|---|---|---|---|---|
7816LR4 |
7812R4 |
7808R4
|
7804R4
|
||
Description | Delivers the highest capacity, non-blocking Ethernet switching system. | Scalable non-blocking capacity for large scale deployments | Delivers high density, low power non-blocking Ethernet switching system. | Optimum form-factor for many medium to large deployments. | |
Switching Capacity | 460 Tbps (920 Tbps FDX) |
345 Tbps (690 Tbps FDX) |
230 Tbps (460 Tbps FDX) |
115 Tbps (230 Tbps FDX) |
|
Linecard Capacity | 28.8 Tbps | 28.8 Tbps | 28.8 Tbps | 28.8 Tbps | |
800G Interfaces | 576 | 432 | 288 | 144 | |
400G Interfaces | 1152 | 864 | 576 | 288 | |
Forwarding Rate | 173 Bpps | 130 Bpps | 86 Bpps | 43 Bpps | |
Total Buffer | 512 GB | 384 GB | 256 GB | 128 GB | |
Rack Units | 32 | 23 | 16 | 10 | |
Airflow | Front to Rear | Front to Rear | Front to Rear | Front to Rear |
7800R4 800G Linecards:
|
|
---|---|
Ports | 36 800G OSFP |
Max 800G | 36 |
Max 400G | 72 |
Max 100G | 288 |
Port Buffer | 32 GB |
Switching Capacity | 28.8 Tbps |
Forwarding Rate | 10.8 Bpps |
Cluster Computing Optimized | Yes |
Combined with the highly modular, robust Extensible Operating System (EOS), a dynamic smart deep buffer architecture with virtual output queuing for lossless forwarding, low latency and wire speed performance, the Arista 7800R4 line cards enable high performance, high scalable solutions for accelerated computing clusters.
Datasheet
Accelerated Computing
- .Arista EtherLink Portfolio
- .Arista's Etherlink™ Smart Networking Portfolio
- .AI Networking White Paper
- .Simplifying and Scaling AI Workloads with VAST and Arista
- .Jayshree's Blog: The Arrival of AI Networking at Petascale
- .Jayshree's Blog: The Next Frontier in AI Networking
- .Jayshree's Blog: The Arrival of Open AI Networking
- .Lossless Network for AI/ML/Storage/HPC with RDMA
- .PF39 RoCE Integrator’s List
- .DDN A3I ® Solutions With Supermicro X12 Gaudi ® AI Servers