mirror of
https://github.com/tailscale/tailscale.git
synced 2025-02-12 12:29:41 +00:00
Two levels of queueing are unnecessary. The resulting implementation performs as follows under request bursts (`count` packets sent concurrently): lost count avg latency 0 / 256 (00.00%) - 28ms 0 / 512 (00.00%) - 146ms 0 / 768 (00.00%) - 166ms 0 / 1024 (00.00%) - 416ms 11 / 1280 (00.86%) - 430ms 145 / 1536 (09.44%) - 715ms 364 / 2048 (17.77%) - 836ms Signed-off-by: Dmytro Shynkevych <dmytro@tailscale.com>