AFTER: Algorithmic Framework for Throughput EstimatoRs for IEEE 802.11 networks

Network Performance Estimator with Applications to Route Selection for IoT Multimedia Applications

“…One notable example of this latter class of performance estimators is AFTER [23]. It was proposed to tackle the problem of real-time throughput estimation for multihop IEEE 802.11 networks.…”

“…Algorithm 1 illustrates MAPE's overall operation, which is divided in three steps: Step 1 -MAPE starts with a complete snapshot of the current network state as input consisting of a representation of the network topology that includes link quality estimates (i.e., link frame delivery probability), list of currently active flows along with the respective paths, and each flow's data rate; Step 2 -MAPE then uses the initial network snapshot to simulate the network as it operates until reaching steady state, which is used to compute long term throughput, packet loss, and end-to-end delay estimates in Step 3. Note that we employ the term steady state in the same sense as in [23], i.e., as a finite cycle of states that repeats themselves. At the end of each iteration, MAPE stores a snapshot of the current network state, which consists of currently ongoing transmissions with their respective remaining times, the content of the queues and the backoff counter of the wireless medium access for all nodes that are traversed by any flow on the evaluated flow set, and the current medium access priority list.…”

“…MAPE's current implementation uses AFTER [23] as the underlying deterministic performance estimator. As shown in Figure 1, MAPE starts by initializing the simulation state with its input arguments.…”

“…We expect MAPE to achieve predictions close to those of ns-3, but in a reproducible manner and at a fraction of the required execution time. We also use SSIM [27] and another metric called classification inversions -as defined in [23], and further explained in Subsection 6.2 -to evaluate video quality and demonstrate the practical suitability of MAPE to the problem of route selection for multimedia applications.…”

“…To evaluate how MAPE can be used to inform path selection for video transmission, we use the concept of classification inversions [23] defined as follows. Consider two different paths a and b.…”