Network-Analysis

Installation Guide

1. Create new Virtual Environment:

virtualenv venv

2. Install requirements:

pip install -r requirements.txt

3. Change db setting to your local settings, e.g:

DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.mysql',
        'NAME': 'djangodatabase',
        'USER': 'dbadmin',
        'PASSWORD': '12345',
        'HOST': 'localhost',
        'PORT': '3306',
    }
}

4. Run migrations:

python manage.py migrate

5. Optionally create a superuser in order to log into the admin site:

python manage.py createsuperuser

System Design

Because the system is mainly a data pipeline that run at equal intervals, the diagram-type chosen to illustrate the design is flowchart:

At the implementation level this pipeline is implemented in two separate modules:

1. kpis/pipeline.py: This module is pythonic (has no dependency on Django), and its responsability is to read the files, one by one so that memory footprint at minimum, parse them and aggregate them in pandas Dataframe.

2. kpis/management/commands/update_network_stats.py: This is a custom Django management command, which initiate the update process and persist the results. Also in this command, a check is performed to see if we should do the one-hour aggregations step.

To run this pipeline, you could set a cronjob inside the crontab, which runs at every 5th minute:

 */5 * * * * cd /home/my/NetworkAnalysis && /home/my/virtual/bin/python /home/my/NetworkAnalysis/manage.py update_network_stats > /tmp/cronlog.txt 2>&1

API Design

1. GET /api/KPIs/

List all the available KPIs, sample response:

[
    {
        "id": 1,
        "name": "ServiceTrafficVolume"
    },
    {
        "id": 2,
        "name": "CellUniqueUsers"
    }
]

2. GET /api/KPIs/<id>/

List the data available for kpi with identifier id. This endpoint is filterbale by three parameters:

• interval_start_timestamp__gte: List records with interval_start_timestamp greater than or equal to the value passed.

• interval_end_timestamp__lte: List records with interval_end_timestamp less than or equal to the value passed.

• interval: Filter records by interval type five-minutes (with value of 1), or one-hour (with value of 2).

Sample response:

Request: GET /api/KPIs/1/?interval_start_timestamp__gte=2021-07-17T23%3A10%3A00Z&interval_end_timestamp__lte=&interval=1

Reponse:

[
    {
        "id": 1,
        "interval_start_timestamp": "2021-07-18T00:00:00Z",
        "interval_end_timestamp": "2021-07-18T00:05:00Z",
        "interval": 1,
        "service_id": 1,
        "total_bytes": 16100
    },
    {
        "id": 2,
        "interval_start_timestamp": "2021-07-18T00:00:00Z",
        "interval_end_timestamp": "2021-07-18T00:05:00Z",
        "interval": 1,
        "service_id": 3,
        "total_bytes": 11500
    },
    {
        "id": 3,
        "interval_start_timestamp": "2021-07-18T00:00:00Z",
        "interval_end_timestamp": "2021-07-18T00:05:00Z",
        "interval": 1,
        "service_id": 2,
        "total_bytes": 9260
    }
]

Notes:

• To ease testing, hour aggregation runs always (you can disable this in kpis/management/commands/update_network_stats.py by removing the settings.DEBUG check).

• Also to ease testing, you should replace the timestamp part of the files names with the current day first interval e.g datetime (July 18, 2021 0:05:00) the timestamp would be 1626566700000.