| NOTE: following tests were done in Mar. 2025, which may not reflect latest status of the package. |
AimRT 0.10.0-py Performance Test Report
Preface
AimRT’s communication layer is implemented via plugins. The official plugins support iceoryx, ROS2, Zenoh, Http, Grpc, Mqtt, and other communication backends, covering common edge and cloud communication scenarios. These plugins provide the two common communication patterns of publish-subscribe and request-response to enable multi-process communication both locally and across machines.
Test Environment
System Environment:
Operating System:
CPU: 13th Gen Intel® Core™ i5-1350P
Linux 6.1.59-rt16
Software Environment
AimRT Version: 0.10.0
Test Items
Tests were conducted using AimRT-py; the test items are as follows:
Single-machine performance tests
Channel backend performance tests
Impact of packet size on performance
Impact of topic count on performance
Impact of concurrency on performance
Rpc backend performance tests
Impact of packet size on performance in bench mode
Impact of packet size on performance in fixed_freq mode
Impact of concurrency on performance in bench mode
Impact of concurrency on performance in fixed_freq mode
Multi-machine performance tests
Channel backend performance tests
Impact of packet size on performance
Impact of topic count on performance
Impact of concurrency on performance
Rpc backend performance tests
Impact of packet size on performance in bench mode
Impact of packet size on performance in fixed_freq mode
Impact of concurrency on performance in bench mode
Impact of concurrency on performance in fixed_freq mode
Test Results
Single-machine Performance Test (X86)
Channel Backend Performance Test
Impact of Packet Size on Performance:
Test Purpose: Performance test of the single-machine cross-process Channel backend under different
packet sizesTest Configuration:
channel_frequency: 1 kHz
pkg_size: 256 B ~ 64 KB (2^8 ~ 2^16, increasing in powers of 2)
topic_number: 1
parallel_number=1
Test Results:
Impact of Topic Count on Performance:
Test Purpose: Performance test of the single-machine cross-process Channel backend under different
topic countsTest Configuration:
channel_frequency: 1 kHz
pkg_size: 1024 B
topic_number: 1 ～ 10
parallel_number=1
Test Results:
Impact of Concurrency on Performance:
Test Purpose: Performance test of the single-machine cross-process Channel backend under different
concurrency levelsTest Configuration:
channel_frequency: 1 kHz
pkg_size: 1024 B
topic_number: 1
parallel_number=1 ～ 10
Test Results:
Rpc Backend Performance Test
Impact of Packet Size on Performance in Bench Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
packet sizesin bench modeTest Configuration:
mode: bench
channel_frequency: 1 kHz
pkg_size: 256 B ~ 64 KB (2^8 ~ 2^16, increasing in powers of 2)
paraller_number: 1
Test Results:
Impact of Packet Size on Performance in Fixed-Freq Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
packet sizesin fixed-freq modeTest Configuration:
mode: fixed-freq
channel_frequency: 1 kHz
pkg_size: 256 B ~ 64 KB (2^8 ~ 2^16, increasing in powers of 2)
paraller_number: 1
Test Results:
Impact of Concurrency on Performance in Bench Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
concurrency levelsin bench modeTest Configuration:
mode: bench
channel_frequency: 1 kHz
pkg_size: 1024 B
paraller_number: 1 ～ 10
Test Results:
Impact of Concurrency on Performance in Fixed-Freq Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
concurrency levelsin fixed-freq modeTest Configuration:
mode: fixed-freq
channel_frequency: 1 kHz
pkg_size: 1024 B
paraller_number: 1 ～ 10
Test Results:
Cross-machine Performance Test
Channel Backend Performance Test
Impact of Packet Size on Performance:
Test Purpose: Performance test of the single-machine cross-process Channel backend under different
packet sizesTest Configuration:
channel_frequency: 1 kHz
pkg_size: 256 B ~ 64 KB (2^8 ~ 2^16, increasing in powers of 2)
topic_number: 1
parallel_number=1
Test Results:
Impact of Topic Count on Performance:
Test Purpose: Performance test of the single-machine cross-process Channel backend under different
topic countsTest Configuration:
channel_frequency: 1 kHz
pkg_size: 1024 B
topic_number: 1 ～ 10
parallel_number=1
Test Results:
Impact of Concurrency on Performance:
Test Purpose: Performance test of the single-machine cross-process Channel backend under different
concurrency levelsTest Configuration:
channel_frequency: 1 kHz
pkg_size: 1024 B
topic_number: 1
parallel_number=1 ～ 10
Test Results:
Rpc Backend Performance Test
Impact of Packet Size on Performance in Bench Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
packet sizesin bench modeTest Configuration:
mode: bench
channel_frequency: 1 kHz
pkg_size: 256 B ~ 64 KB (2^8 ~ 2^16, increasing in powers of 2)
paraller_number: 1
Test Results:
Impact of Packet Size on Performance in Fixed-Freq Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
packet sizesin fixed-freq modeTest Configuration:
mode: fixed-freq
channel_frequency: 1 kHz
pkg_size: 256 B ~ 64 KB (2^8 ~ 2^16, increasing in powers of 2)
paraller_number: 1
Test Results:
Impact of Concurrency on Performance in Bench Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
concurrency levelsin bench modeTest Configuration:
mode: bench
channel_frequency: 1 kHz
pkg_size: 1024 B
paraller_number: 1 ～ 10
Test Results:
Impact of Concurrency on Performance in Fixed-Freq Mode:
Test Purpose: Performance test of the single-machine cross-process Rpc backend under different
concurrency levelsin fixed-freq modeTest Configuration:
mode: fixed-freq
channel_frequency: 1 kHz
pkg_size: 1024 B
paraller_number: 1 ～ 10
Test Results:
