|
Embedded Linux Realtime Application Performance Engineering |
|
A large, parallel, gigabit Ethernet computing architecture using 8-core processors with embedded realtime Linux required performance engineering in order to; 1) understand the realtime performance of the Linux 2.6 kernel as built for the 8-core processors, 2) understand the realtime performance of the application software and 3) tune the Linux 2.6 kernel and the application software in order to achieve maximum efficiency. Static and dynamic analysis techniques were used that leveraged open source tools to profile the software such as valgrind, cppcheck, wireshark, oprofile, and collectd. Additional collectd plugins were developed to collect application specific information. We've developed significant core competencies in the performance engineering of complex, embedded realtime Linux systems.
|