AdpGen, a workload generator for live events broadcasted through HTTP Adaptive Streaming. It can be used, for instance, to generate a synthetic workload in order to evaluate server resource consumption.
- Based on 2014 FIFA World Cup event: https://ieeexplore.ieee.org/document/8969602/
- Generates the client on-time, inter-session time, number of sessions and adaptation regime
To call the generator from the command line use:
python generator_cluster_markov_thesis_v2.py [execution round number] [random seed];
Example: python generator_cluster_markov_thesis_v2.py 1 1276;
The output is stored in the synthetic folder:
- cluster_ontime_x.csv: session duration in seconds
- cluster_offtime_x.csv: client inter-session duration
- cluster_qtd_x.csv: client number of sessions
- cluster_other_x.csv: performance metrics (average bitrate, number of adaptations)
- cluster_client_enter_x.csv: client arrival time
Number of clients
To change the number of clients, change the value of variable TOTAL_CLIENTS on line 136
Arrival regime
Client arrival follows a uniform distribution where probabilities are stored in the file probabilidades_arrival_global_14400.csv
To change the file as well as the simulation lengh, please manipulate the following instruction user_arr = np.random.choice(a = np.arange(1,14401,1), size=clients_list, replace = True, p = read_csv("probabilidades_arrival_global_14400.csv"))