ABNF codec of SIP usually accounts for 40%-60% of CPU performance
consume of the entire protocol software applications,. while the
comparability of performance test of the codec module is relatively strong.
Hereby, we compare these two as an example.
• Test conditions: the same SIP signaling and hardware (Celeron mobile 1.3G,
384 RAM). Besides, the settings of codec of Juphoon SIP are all
meticulous parsing.
• Test samples: the 13 samples are INVITE, 180,183,200, ACK, BYE, INFO,
PRACK, INVITE (ISUP), 200 (BYE), 200 (INFO), 200 (PRACK1) and 200
(PRACK2). Their packet length is 5150 byte totally, the number of header
field is 109, the processing repetition is 30, 000 times and the time unit is
second.
Compared with a number of commercial SIP protocol stacks of professional protocol
providers, the average parsing of Juphoon SIP protocol stack is 3-9 times faster. Hence, we conclude that the
performance of Juphoon SIP protocol stack is relatively prominent.

From left to right is the times that TSIP, OSIP, RSIP and Juphoon
SIP used for repeatedly encoding and decoding.