Introducing Virgo Network, Google’s scale-out AI data center fabric

Google introduces Virgo Network, a megascale AI data center fabric addressing bandwidth and latency constraints of traditional networks.

• •Three-layer architecture: scale-up domain for pod interconnects, scale-out RDMA fabric for accelerator communication, and Jupiter network for storage/compute access
• •Supports 134,000 TPU chips with up to 47 petabits/sec non-blocking bandwidth and 40% lower latency than previous generations
• •Flat two-layer non-blocking topology using high-radix switches reduces network layers and enables deterministic low latency
• •Independent switching planes and sub-millisecond telemetry enable fault isolation and rapid straggler detection

This summary was automatically generated by AI based on the original article and may not be fully accurate.