Specifications
Model
Edgecore DCS560
Total switching capacity
51.2 Tbps
Port configuration
64 ports supporting 800G interfaces
Switch silicon
Broadcom StrataXGS Tomahawk 5 series
Form factor
2RU
Interface variants
OSFP800 optical interfaces or QSFP-DD800 interfaces
Optics support
Copper cables for short reaches plus long-distance ZR+ pluggable optics
Network operating system
Open-source SONiC
Reliability features
Redundant power supplies and redundant fans
Availability target
Five-nines (99.999%) mission-critical deployments
Operating temperature
Wide operating temperature range
Traffic features
Adaptive routing and dynamic load balancing for low-entropy, bursty AI/ML flows
Generational gains
2x bandwidth capacity and lower power per bit versus the previous generation
Roadmap
Higher-capacity additions planned for the AI/ML use case
Overview
The Edgecore DCS560 is an 800G-optimised data-centre switch aimed at AI and machine-learning fabrics. A single 2RU chassis delivers 51.2 Tbps of switching capacity across 64 800G-capable ports using Broadcom's StrataXGS Tomahawk 5 silicon, offered with either OSFP800 or QSFP-DD800 interfaces so operators can pick the optic and cable ecosystem their site already uses.
AI training and inference traffic is described by Edgecore as low-entropy, high-intensity and bursty, which is exactly the pattern that defeats static hash-based load balancing. The DCS560 counters with adaptive routing and dynamic load balancing to shorten job completion times. It ships with redundant power supplies and fans for five-nines deployments and runs the open-source SONiC network operating system.
Key Benefits
51.2 Tbps in 2RU concentrates a fat-tree spine tier into fewer chassis. OSFP800 or QSFP-DD800 options avoid re-cabling an existing site. Adaptive routing and dynamic load balancing target the bursty flow profiles AI workloads produce. SONiC keeps the switch in an open, automatable software stack. Redundant PSU and fans suit mission-critical fabrics.
Applications
AI/ML training and inference fabrics, GPU cluster spine and uplink tiers, hyperscale leaf-spine networks, storage back-end fabrics, and telecom cloud deployments that standardise on SONiC.
Request a Quote — EDGECORE DCS560 — 51.2 TBPS 800G AI/ML DATA CENTRE SWITCH WITH SONIC
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →