Cool asymmetric network performance happened to mess up a customer benchmark

A bunch of Unison systems, a 40 GbE network interconnecting them, and a bunch of client nodes on 40GbE -> 4x 10GbE links (to accomodate enough clients for the load testing). 40GbE < -> 40GbE works great. Full bandwidth, only minor oddities (single thread performance around 27Gb/s, need multiple threads to hit 40). 10GbE < -> 10GbE works great. Full bandwidth, nothing odd. 10GbE -> 40GbE works great, get about the expected bandwidth (10GbE). Writes work well. 40GbE -> 10GbE is sorta kinda broken. We get anywhere from 170Mb/s to 1.5Gb/s. My switch FW is badly out of date, so its time to upgrade. Checking in to see if this is a known problem, but … wow … wasn’t expecting that. Might need to see if I can drop a few clients and go all 40GbE. If so, we’ll do that for the moment.