Arista 7280R3 / 7800R3 Series
The Arista 7280R3 and modular 7800R3 series stand among the deepest-buffered merchant silicon platforms available. Leveraging Broadcom Jericho2 architecture, they are positioned for IP storage, global WAN transit, and content delivery networks where absolute lossless delivery is necessary.
| Field | Value |
|---|---|
| Ports | 10G/25G through 400G scale |
| Minimum Fast Buffer | 32 MB on-chip per ASIC |
| Deep Buffer (HBM) | Up to 24 GB / 32 GB off-chip per chassis subsystem |
Buffer Architecture
The 7280R3 utilizes an ultra-deep buffer structure enabled by Virtual Output Queuing (VOQ) and High Bandwidth Memory (HBM) integration natively with Jericho2.
The platform separates traffic dynamically:
- Short-lived flows use the 32 MB on-chip fast memory.
- Large microburst flows and congestion events are seamlessly shunted into the off-chip HBM pool (ranging from 16 GB to 32 GB depending on model scale).
This mechanism inherently averts head-of-line blocking while maintaining nearly 100% capacity transmission on outbound links without packet degradation. Because congestion is localized and buffered before crossing internal fabrics or facing congested outgoing ports, applications experience pure deterministic routing.