Module nccl_priority 

NCCL Network Priority via eBPF/tc

When the scheduler detects NcclCollective phase, this module uses Linux traffic control (tc) with eBPF classifiers to give NCCL traffic absolute priority on network interfaces.

This reduces all-reduce tail latency which is the critical path in distributed training — the slowest rank determines throughput.

Constants

DEFAULT_INTERFACES 🔒

Network interface to apply NCCL priority rules to.

NCCL_DSCP 🔒

DSCP marking for high-priority NCCL traffic.

NCCL_PORT_END 🔒

NCCL_PORT_START 🔒

NCCL uses a specific port range (typically 29500-29999 for PyTorch).

Functions

detect_nccl_interface

Detect the primary network interface for NCCL traffic.

disable_nccl_priority

Remove NCCL priority rules.

enable_nccl_priority

Apply high-priority tc rules for NCCL traffic. Requires root / CAP_NET_ADMIN.

mark_nccl_dscp

Mark outgoing NCCL packets with DSCP EF for switches/routers.