udplogd has been developed to deal with large amounts of
logging messages, which cannot be handled by other logging
demons properly. To achive this, udplogd has been stripped
of all unecessary add-ons like name resolving, priorities
and facilities. It just concatenates lines of raw UDP datagrams.
udplogd has been developed under Solaris to handle large
amounts of logging information being delivered by several cascaded
internet routers. Although syslog's host name resolving mechanisms
can be cached, udplogd proved to work much more efficiently
than the former. Being provided with log rotation facilities
of its own it also packed the data into several handy files,
each of which could be browsed comfortably and independently.
udplog's second operational area was its duty to deal with
the log files of an ISP's line-up of mail busy servers logging
to a StorEdge RAID-array being connected to a Sun E250
dual processor machine. Together the mail servers yielded
a more or less continuous data rate of about 20 KB/s
mail server logging messages.
syslogd was unable to deal with this data rate
in the long run. udplogd, on the other hand, managed
to receive and log the data without causing too much system
load. The client mail servers logged their data to a
named pipe locally from which sendudp(1) fetched it and
transmitted it to the central logging server.
Note, that udplogd should not be used to transfer files
of important content or
binary data over networks. Since it is based on UDP datagrams,
its transfer being unreliable comes natural.