HP-UX TCP/IP Performance White Paper, March 2008

socket_qlimit_max:

Sets maximum number of connection requests for non-AF_INET sockets.

[1-2147483647] Default: 4096

socket_udp_rcvbuf_default:

Sets the default receive buffer size for UDP sockets.

The value of this tunable parameter should not exceed the

value of the ndd parameter udp_recv_hiwater_max. Otherwise

a socket() call to create UDP socket will fail and return

the errno value EINVAL. [1-2147483647] Default: 65535

socket_udp_sndbuf_default:

Sets the default send buffer size for UDP sockets.

[1-65535] Default: 65535

TCP Tunables

tcp_conn_request_max:

Maximum number of outstanding inbound connection requests.

[1, - ] Default: 4096 connections

This is also known as the maximum depth of the "listen queue." The actual maximum for any given TCP

endpoint in the LISTEN state will be the MINIMUM of tcp_conn_request_max and the value the

application passed-in to the listen() socket call.

For this parameter to take effect for a given application, it must be set before said application makes its call

to listen(). So, if you use ndd to set this value after the application has started, it will have no effect

unless you can get the application to recreate its LISTEN endpoint(s) typically by stopping and restarting the

application.

You can see if tcp_conn_request_max might be too small by looking at the output of either netstat

-s or netstat -p tcp and looking for the line displaying the number of connection requests dropped

due to full queue. If the number of drops is zero, the value of tcp_conn_request_max is fine. If the value

is non-zero, either tcp_conn_request_max is too small, or the values the applications are using in their

calls to listen() are too small.

Setting this value higher has no particular downside especially for any "internet server" system.

tcp_cwnd_initial:

RFC 2414 defines a formula for calculating the sender's initial

congestion window that usually results in a larger window than in

previous releases.

The default initial congestion window is now calculated using the