Straggler Problem

Appears in 1 paper

When one GPU is slower than others (older hardware, thermal throttling, interference), it becomes the bottleneck.

As used in Paper 19 — Ring Attention with Blockwise Transformers for Near-Infinite Context →

When one GPU is slower than others (older hardware, thermal throttling, interference), it becomes the bottleneck. The entire ring must wait for the straggler to finish each round. One slow GPU reduces effective speedup from P× to less.