VoIPER Performance Testing

From VoIPER

This page details the results of reliability and performance tests performed using VoIPER. The aim is to establish if any reliabilty, memory or other performance issues exist in the dev branch of VoIPER.

Contents 1 Test #1.1: Protocol based crash detection, SDP fuzzer, dev branch 2 Test #1.2: Protocol based crash detection, SDP fuzzer, stable branch 3 Test #2.1: Process based crash detection, SDP fuzzer, dev branch 4 Test #2.2: Process based crash detection, SDP fuzzer, stable branch 5 Backend testing without fuzzer

Test #1.1: Protocol based crash detection, SDP fuzzer, dev branch

RSIZE begin: 47mb

RSIZE end:

Time to replay 5000 fuzz requests:

Requests per second:

False positives:

Exceptions:

Number of transactions where no response was received:

Number of unexpected responses:

Test #1.2: Protocol based crash detection, SDP fuzzer, stable branch

Time to replay 5000 fuzz requests:

Requests per second:

False positives:

Test #2.1: Process based crash detection, SDP fuzzer, dev branch

RSIZE begin:

RSIZE end:

Time to replay 5000 fuzz requests:

Requests per second:

False positives:

Exceptions:

Number of transactions where no response was received:

Number of unexpected responses:

Test #2.2: Process based crash detection, SDP fuzzer, stable branch

Time to replay 5000 fuzz requests:

Requests per second:

False positives:

Backend testing without fuzzer

• Approx. 4 full INVITE sequences inc. a CANCEL, per second against Asterisk (21 seconds for 100 requests). The transactions consisted of an INVITE, waiting for a 1XX provisional, a CANCEL, waiting for a 3XX-6XX and an ACK to that.
• This is reduced to just under 3 per second if a REGISTER is sent before every INVITE (34 seconds for 100 requests)
• Including the fuzzer reduces this to 2.3 requests per second (128 seconds for 300 requests)