An Exposed Closed-Loop Model forCustomer-Driven Service Assurance Automation

This repository accompanies the conference paper with the same title: "An Exposed Closed-Loop Model forCustomer-Driven Service Assurance Automation", published in EuCNC 2021. In the paper we propose an exposed CL model based on service exposure which is meant to automate service assurance tasks like auto-scaling in a network function virtualization (NFV) system orchestrated by ETSI Open Source MANO (OSM). Here we provide the testbed-generated experimental results we used to validate our proof-of-concept.

Please refer to the conference paper for details about the topology and the scenario of each experiment.

The csv files can be grouped in 3 categories, which correspond to the three vantage points (levels) of monitoring. Each experiment repetition is represented by a hash. The hash is either in the file name or as a column inside the csv files.

1. prometheus.csv

Some of the columns are:

• timestamp: timestamp in Oslo timezone
• experimentHash: the unique identifier of the experiment
• vdu_name: Where the data comes from
• metric_name: what is being measured
• value: the actual value.

It was hard to make this csv flat (ie. use as a unique index the timestamp), because the column names would be very long. The way to use it is to filter per, <experimentHash, vdu_name, metric_name> and then use <timestamp, value> to plot the timeseries of the value for this metric (experimentHash) of this component (vdu_name) for this experiment (experimentHash).

2. sysstat csv

There are 3 csv with the systat values . They contain exactly the same information just in different representations. So, pick the one that is most convenient for you.

3. packet capture csv (csv that starts with a hash)

These are csvs from the packet capture (pcap) files. You can tell by the name which component created them. The files that end with a number are captured at the client (UE). Refer to the topology figure of the conference publication.