Yet Another Centralized Scheduler

                                                                                     
YYYYYYY       YYYYYYY           AAA                  CCCCCCCCCCCCC   SSSSSSSSSSSSSSS 
Y:::::Y       Y:::::Y          A:::A              CCC::::::::::::C SS:::::::::::::::S
Y:::::Y       Y:::::Y         A:::::A           CC:::::::::::::::CS:::::SSSSSS::::::S
Y::::::Y     Y::::::Y        A:::::::A         C:::::CCCCCCCC::::CS:::::S     SSSSSSS
YYY:::::Y   Y:::::YYY       A:::::::::A       C:::::C       CCCCCCS:::::S            
   Y:::::Y Y:::::Y         A:::::A:::::A     C:::::C              S:::::S            
    Y:::::Y:::::Y         A:::::A A:::::A    C:::::C               S::::SSSS         
     Y:::::::::Y         A:::::A   A:::::A   C:::::C                SS::::::SSSSS    
      Y:::::::Y         A:::::A     A:::::A  C:::::C                  SSS::::::::SS  
       Y:::::Y         A:::::AAAAAAAAA:::::A C:::::C                     SSSSSS::::S 
       Y:::::Y        A:::::::::::::::::::::AC:::::C                          S:::::S
       Y:::::Y       A:::::AAAAAAAAAAAAA:::::AC:::::C       CCCCCC            S:::::S
       Y:::::Y      A:::::A             A:::::AC:::::CCCCCCCC::::CSSSSSSS     S:::::S
    YYYY:::::YYYY  A:::::A               A:::::ACC:::::::::::::::CS::::::SSSSSS:::::S
    Y:::::::::::Y A:::::A                 A:::::A CCC::::::::::::CS:::::::::::::::SS 
    YYYYYYYYYYYYYAAAAAAA                   AAAAAAA   CCCCCCCCCCCCC SSSSSSSSSSSSSSS   

Layout

Code

• master.py and master_utils
• worker.py
• config_utils/: utilities related to reading the config.json file which defines workers.
• job_utils/: definitions of tasks, jobs and messages shared by master and workers.
• extras/: extra scripts such as config generator, sending requests to master, etc.
• start_workers.py: starts all workers defined in config.json locally with the correct ports. Workers are stopped once the script exits.

Documentation

Documentation - report and architecture

Analysis

Analysis

Usage

• Edit the config.json file with correct worker details. Or generate one.

• First the master

  python3 master.py # uses the random scheduler
  

  OR

  python3 master.py rr # uses the round robin scheduker
  

  OR

  python3 master.py ll # uses the least loaded scheduler
  

• Now the worker

  python3 start_workers.py # uses the config.json
  

• Now run extras/requests.py with the number of requests you want to send to master node. It should send and finish. Let it run in the background. Example:

  python3 extras/requests.py 100
  

  For more customized testing, requests_eval.py

  python3 extras/requests_eval.py # custom task generated
  

Analysis

Copy over the required logs to analysis/.

cd analysis
python3 yacs_logs_analysis.py # task 1
python3 graph_task2.py # task 2

Non-local workers

The external-workers branch has experimental support for non-local workers. Tested on a Raspberry Pi node.