- Added reassignment types "max_prop" and "always_attend" to FCFS policy.
- Added thread that reads request file.
- Add requests file.
- The arrival of requests are visualized in the Gantt chart.
- Structural changes to scheduling_policy.
- Added Request class.
- Code modularization in scheduler.
- Added docker pause handling.
- Added active container handling when you want to finish the scheduler.
- Add new event (PAUSE_CONTAINER and RESUME_CONTAINER).
- Fixed problem in gantt chart plotting.
- Add plotly and pandas libraries.
- Request queues now belong to the scheduling policy.
-
Add support for Gantt Charts.
-
Scheduler host:
- Changes in update function to return parallelism applied to container
- Add event logs for Gantt chart.
- Fixed issue in the choice of container to generate an update request.
- Fixed issue in docker container prune which was running on every message wait.
- Add event values in CSV file.
-
Remove finished container with command "docker container prune -f"
-
Fix problems in event logs.
-
Attribute error in client: the signal variable had the same name as the signal class.
-
Add JSON objects messages over socket.
-
Trace plotting problem:
- Add the call to add_thread in the creation of the main threads.
- Fixed problem in the amount of thread plots.
-
Sorting the folders
- Add Commons flder. Includes log files for the entire project.
- Change name of dockerfile.
-
Change dockerfile to indicate how the tensorflow program should be entered.o
-
Scheduler
- Correction in the generation of events.
-
Client:
- Correction in the generation of events.
- Add number of thread parameter in signal handler.
- Correction socket messages.
-
Trace:
- Add the registration of a new thread to save its execution traces.
- Add second value in trace event.
- Thread index correction.
-
The scheduler correctly responds to the execution and update requests.
-
Scheduler:
- Add keyboard input for the completion of the program.
- Add the joins of the client threads.
- Changes in waiting times to generate the next execution or update request.
- Fix storing scheduler events.
- Code ordering.
-
Client:
- The reception of messages from the scheduler was centralized in a single thread.
- Changes in the handling of global variables.
-
Trace:
- You can indicate the number of events drawn in the plot_events function.
-
Add events class. Stores the events that can be saved in the execution traces.
-
Execution info:
- Add the attribute container number.
- Scheduler:
- Error was resolved in evaluated condition (Attention thread).
- Add evaluation of the status of the current request when old requests are served.
- Modification of the place where the host port is increased for the next execution request.
-
Scheduler:
- Changes in update request processing. Receives available resources and should not query the system.
- Add error handling in the search for the container you want to change (Update Request thread).
- Changes in the generation of execution requests (Execution Request thread).
- Changes in the reception of client data.
- Changes in the function that reallocates resources in actives containers.
- Changes in attention of execution/update requests (Attention thread).
-
Client:
- Changes in the signal handler of the Tensorflow program. Advises the client to finish
- Change in the attention of update requests. Condition when parallelism update runs.
- Changes in the reception of data through the socket. Now a single data is received and not an array of 1024.
- Add error handling in main program.
-
Execution info:
- Changes in parallelism update.
-
Scheduling policy:
- Changes in parallelism schedule for policies FFSnotReassigment and FFSwithReassigment.
-
Scheduler:
- Changes in update request processing.
- Change in the creation of execution requests.
- Changes in the attention of execution/update requests.
- Changes to screen printed texts.
-
Client:
- Add logging library to store tracking information in a text file.
- Correction of errors in the generation of strings.
-
Trace:
- The function to store the traces in CSV format now receives the directory by parameter.
-
System Info:
- Add information about total number of cores in the machine/server.
- Add client executed in the container that allows to control the execution of the Tensorflow program.
- Add scheduler. Recieves requests for execution and update of tensoroflow programs. The requests have the following data: *type of request *container name *docker image *inter parallelism *intra parallelism
- Add trace class. Allows you to store events from the scheduler or client, plot the events in a graph and store them in a file with CSV format.
- Add system class. Stores the information on the amount of resources available in the machine.
- Add scheduling policy class. Defines the planning policies of the execution and update requests.
- Add execution info class. Define information of an active container.
- Project creation.