HP-UX 11i Release Notes (December 2000)

Internet and Networking Services
Network Transport (new)
Chapter 10196
setsockopt or t_optmgmt can set for a TCP socket. These systemwide
parameters prevent processes from keeping large amounts of data in
send or receive buffers and consuming system resources.
tcp_fin_wait_2 is used to set how long a connection will be in
FIN_WAIT_2. Use this cautiously. If the remote TCP entity is slow, but
would terminate normally (is not hung nor will it terminate abnormally),
TCP may close the connection prematurely. A possible side effect is that
data in the remote connections receive buffer may be flushed. If this
happens unexpectedly, then the data could become corrupted.
The above description is an enhancement.
ip_udp_status has a new field. It now will report how many times a
given UDP socket has overflowed. This only works on sockets currently
open. This is a very handy troubleshooting tool used when netstat -p
udp shows socket overflows. This feature is an enhancement.
4) TimeStamps option supported: tcp_ts_enable allows RFC 1323
TimeStamp extensions to TCP Headers. The TimeStamps are used for
two purposes:
1. RTTM (Round Trip Time Measurement): Interval between time TCP
sends a segment and the time return acknowledgement arrives.
2. PAWS (Protect Against Wrapped Sequences) on high-speed networks.
Supported parameter values are:
2 Use Timestamps option if initiated by the remote
system
1 Always try to initiate the use of Timestamps option
0 Never use Timestamps option
5) Socket structure caching can increase performance:
tcp_conn_strategy enables socket caching. It sets how many cached
socket structures the system keeps. The default value of 0 (zero) disables
the feature. A value between 1 and 512 sets a minimum of 512. Any
number above 512 sets tcp_conn_strategy to that value. Enabling socket
structure caching can increase system performance if there are many
short-lived connections on the system.
6) Initial TCP congestion window size is now configurable:
tcp_cwnd_init sets the sender’s initial congestion window size
according to the following formula: