As I understand it, the TCP or UDP recv_hiwat tunable should be roughly equal to the bandwidth delay product ?
For a 100Mbps link I can see that a setting of 64KB should be plenty, but what if you have a 1Gbps link with similar latency, should I tune the tcp_recv_hiwat up to around the 250KB mark? (I understand the both the increased impact this has on the required memory for the buffers and that it only relates to RFC1323-compliant systems. I'm asking because I'm seeing NoCanPuts on a 1Gbps interface which has already had the TCP and UDP recv_hiwat set to 64KB) If this is the case, is it possible to tune this variable for 100Mbps and 1Gbps links seperately ? This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
