Can anybody help me out here. I've checked what I can, but I'm still relatively new to Solaris and I'm stumped as to what's going on here.
I have two Solaris machines connected via gigabit ethernet, but it looks like network transfers between the two are limited to around 10MB/s. I first spotted it doing a zfs send / receive, but since then have tried using mbuffer with similar results. It's as if the network is running at 100Mb/s, but both the machines and the switch appear to have everything running at gigabit speeds. On both machines, solaris believes they are running as full duplex gigabit: server 1 # dladm show-linkprop rge0 LINK PROPERTY VALUE DEFAULT POSSIBLE rge0 speed 1000 -- -- rge0 duplex full -- half,full server 2 # dladm show-linkprop nge0 LINK PROPERTY VALUE DEFAULT POSSIBLE rge0 speed 1000 -- -- rge0 duplex full -- half,full The switch is a 3com 2948-SFP Plus, it reports that both ports have auto-negotiated to full duplex gigabit, and are currently running at 9% of capacity. I tried to use mbuffer to eliminate disk bottlenecks and test the raw network speed with: server 1 (receiving): /usr/local/bin/mbuffer -I server2:10001 -s 128k -m 512M > /dev/null server 2 (sending): cat /dev/zero | /usr/local/bin/mbuffer -s 128k -m 512M -O server1:10001 That averaged around 10MB/s with a peak rate of just over 11MB/s. What else should I be looking at to find out why I'm getting such poor performance between these servers? thanks, Ross -- This message posted from opensolaris.org
