The best practice is to use a syslog aggregation tier and then use the Universal Forwarder on top of your syslog server(s). This gives you the most reliability, auto-load balances the data if you have a distributed setup and will not cause data loss if you need to re-start your Splunk infrastructure. I highly recommend you go this route.
If you don't have syslog servers then you can syslog directly to Splunk.
http://docs.splunk.com/Documentation/Splunk/6.4.1/Data/Monitornetworkports
http://docs.splunk.com/Documentation/Splunk/6.4.1/Data/HowSplunkEnterprisehandlessyslogdata
... View more