The Effective Link Capacity (ELC) metric is a new link quality metric for wireless networks for estimating maximum achievable link capacity. The ELC value of a link represents the achievable throughput a link can reach given the current prevailing link conditions. The metric is able to calculate this value from existing data traffic to aid in decision making such as routing protocols and rate adaptation algorithms. It gathers data either by inspecting outgoing frames itself, or by querying this data from the host kernel.
We describe, design and implement the ELC metric as a Linux user-level process and evaluate the estimation accuracy on a conducted testbed. We find that ELC evaluates the Link Throughput with great accuracy. Further explorations lead us to suggest a new model, which we present as the Model-Based Effective Link Capacity (mELC) metric. While evaluating mELC, we find that non-standard behaviour apparent in our commodity wireless cards, thereby decreasing the estimation accuracy of mELC. Once we account for these non-standard behaviours, we find that mELC increases its estimation accuracy compared to ELC. Unfortunately, the modifications are impractical for general use, and we conclude that the ELC metric is the best trade-off between accuracy and practicality for estimating link throughput.
Finally, we present initial explorations into applications of ELC, including as an improvement in the Minstrel rate-adaptation algorithm. We that the modified Minstrel algorithm RCELC which uses ELC is able to choose greater-throughput links when compared to the original. We also explore using ELC as a replacement for AirTime in the IEEE 802.11s Hybrid Wireless Mesh Protocol. In initial investigations, we find that ELC significantly outperforms two different implementations of AirTime. Both of these works suggest great applications for ELC in the future.