Co-TOMS (Co-optimizing Task Offloading, Memory placement, and voltage Scaling)

This project performs real-time task scheduling based on steady-state genetic algorithms in order to save power consumptions in CPU, memory, and network subsystems with deadline constraints.

Co-TOMS considers three energy-saving techniques, DVFS (dynamic voltage/frequency scaling), hybrid memory placement, and task offloading to edge servers, across different system layers.

Two executables included in this project, which can simulate Co-TOMS in comparison with DVFS, Offloading, and basic configurations.

• gasgen: task generation tool based on CPU and total utilization
• gastask: scheduling scheme generator based on GA

For comparison purposes, our basic simulator supporting dynamic voltage scaling (DVS) and hybrid memory (HM) can be downloaded at https://github.com/oslab-ewha/simrts.

Build

To build gastask and gasgen, use CMake:

$ mkdir -p build && cd build
$ cmake ..
$ make

Run

• Create a new configuration file. Refer to gastask.conf.tmpl.
• run gasgen

# ./gasgen gastask.conf

• Tasks list will be generated into task_generated.txt network_generated.txt network_commander_generated.txtaccording to gastask.conf
• paste task_generated.txt into the task section of gastask.conf
• paste network_generated.txt into the network section of gastask.conf
• paste network_commander_generated.txt into the net_commander_ section of gastask.conf
• run gastask

# ./gastask gastask.conf

• scheduling information is generated in task.txt, which can be used as an input to simrts.

Batch run

• run.sh performs all procedures in batch
  • Before executing run.sh, ./tmp folder should be generated in root.

Data Set

There are two types of data set to perform the simulations of Co-TOMS.

• Synthetic workload:

  • 20% CPU Utilization Workload:
  • 30% CPU Utilization Workload:
  • 40% CPU Utilization Workload:
  • 50% CPU Utilization Workload:
  • 60% CPU Utilization Workload:
  • 70% CPU Utilization Workload:
  • 80% CPU Utilization Workload:
  • 90% CPU Utilization Workload:

• Realistic worload:

  • Robotic Highway Safety Marker (RSM): real-time task set for the actions of mobile robots that carry safety markers in a highway for road construction safety
  • IoT: real-time task set for the actions of a controller in industry machine hands