| | 1 | = Multi Processor Platforms Guide = |
| | 2 | |
| | 3 | This guide describes how to run ALOE in a multi-processor environment in Linux. |
| | 4 | |
| | 5 | Currently, only tcp/ip interfaces are supported, which lack of any QoS mechanism. Therefore, it may produce real-time violations due to transmissions. To minimize this effect, make sure network load is low. |
| | 6 | |
| | 7 | 1) Decide your network topology: number of processors and connectivity. |
| | 8 | 2) Configure External Interfaces Configuration File for each processor (ALOELinuxGuide#ConfigureyourPlatform). Besides data interfaces, you have to configure control and synchronization interfaces. |
| | 9 | 3) Configure Platform Configuration File and select the daemons to launch on every processor. A default configuration maps all manager daemons to a single processor (the one where executable repositories are available) while the rest have only the mandatory daemons: |
| | 10 | * Manager Processor Daemons: hwman, swman, statsman, sync_master, swload, frontend, stats, bridge, exec |
| | 11 | * Rest of Processor Daemons: swload, frontend, stats, bridge, exec, sync |
| | 12 | 3) Launch '''runph''' on every processor |
| | 13 | 4) You will see in the Manager Processor how hwman daemon automatically detects the network of processors. |
| | 14 | 5) Load and run your application normally from the Manager Processor cmdman shell (or from the GUI) |
| | 15 | 6) You can check with execinfo command the actual mapping and execution information of processes. |
| | 16 | |
| | 17 | You should see how the set of modules are mapped to the set of processors according to components resource consumption (defined in application's *.app file) and processors' computing capacity (defined in Platform's Configuration File). |
| | 18 | |
