nqos_split Adapter

class network_gym_client.envs.nqos_split.Adapter(config_json)

nqos_split env adapter.

Parameters:

Adapter (network_gym_client.adapter.Adapter) – base class.

Initialize the adapter.

Parameters:

config_json (json) – the configuration file

Methods

network_gym_client.envs.nqos_split.Adapter.get_observation(self, df)

Prepare observation for nqos_split env.

This function should return the same number of features defined in the get_observation_space().

Parameters:

df (pd.DataFrame) – network stats measurement

Returns:

spaces – observation spaces

network_gym_client.envs.nqos_split.Adapter.get_reward(self, df)

Prepare reward for the nqos_split env.

Parameters:

df (pd.DataFrame) – network stats

Returns:

spaces – reward spaces

network_gym_client.envs.nqos_split.Adapter.get_policy(self, action)

Prepare policy for the nqos_split env.

Parameters:

action (spaces) – action from the RL agent

Returns:

json – network policy

Reward Method

network_gym_client.envs.nqos_split.Adapter.netowrk_util(self, throughput, delay, alpha=0.5)

Calculates a network utility function based on throughput and delay, with a specified alpha value for balancing. Default Reward function.

Parameters:

• throughput – a float representing the network throughput in bits per second

• delay – a float representing the network delay in seconds

• alpha – a float representing the alpha value for balancing (default is 0.5)

Returns:

a float representing the alpha-balanced metric

Additional Methods

network_gym_client.envs.nqos_split.Adapter.get_action_space(self)

Get action space for the nqos_split env.

Returns:

spaces – action spaces

network_gym_client.envs.nqos_split.Adapter.get_observation_space(self)

Get the observation space for nqos_split env.

Returns:

spaces – observation spaces