Estimation of the parameters of token-buckets in multi-hop environments

• Published in: Computer networks (Amsterdam, Netherlands : 1999) Vol. 138; pp. 177 - 191
• Main Authors: Ramos, Javier, Muelas, David, López de Vergara, Jorge E., Aracil, Javier
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 19.06.2018; Elsevier Sequoia S.A
• Subjects: Algorithms; Buckets; Computer networks; Measurement techniques; Network and systems monitoring and measurements; Parameter estimation; Performance evaluation; Quality of service; Quality-of-service management; Service level agreements; Shapers; Taxonomy; Token-Bucket; Token-Bucket; Network and systems monitoring and measurements; Service-level agreements; Quality-of-service management
• ISSN: 1389-1286; 1872-7069
• DOI: 10.1016/j.comnet.2018.04.003

Bandwidth verification in shaping scenarios receives much attention of both operators and clients because of its impact on Quality of Service (QoS). As a result, measuring shapers’ parameters, namely the Committed Information Rate (CIR), Peak Information Rate (PIR) and Maximum Burst Size (MBS), is a relevant issue when it comes to assess QoS. In this paper, we present a novel algorithm, TBCheck, which serves to accurately measure such parameters with minimal intrusiveness. These measurements are the cornerstone for the validation of Service Level Agreements (SLA) with multiple shaping elements along an end-to-end path. As a further outcome of this measurement method, we define a formal taxonomy of multi-hop shaping scenarios. A thorough performance evaluation covering the latter taxonomy shows the advantages of TBCheck compared to other tools in the state of the art, yielding more accurate results even in the presence of cross-traffic. Additionally, our findings show that MBS estimation is unfeasible when the link load is high, regardless the measurement technique, because the token-bucket will always be empty. Consequently, we propose an estimation policy which maximizes the accuracy by measuring CIR during busy hours and PIR and MBS during off-peak hours.