Andrei Gurtov, Sally Floyd, “Modeling Wireless Links for Transport Protocols,” ACM SIGCOMM Computer Communications Review, Volume 34, Number 2, (April 2004). [PDF]
Summary
Intrinsic characteristics of wireless links (e.g., variable bandwidth, bit corruption, channel allocation delays, and asymmetry) significantly affect performance of transport protocols. In addition, different types of wireless links show noticeable variance in different characteristics. This paper summarizes the authors’ experiences in creating realistic simulation models for three different types of wireless links (cellular, WLAN, and satellite links) to evaluate the effect of link level mechanisms on end-to-end transport protocols.
The authors point out that most of the existing models are either not realistic enough to capture the characteristics of different types of wireless links; or realistic but fails to cover large parameter space required to better understand different characteristics; or overly realistic and ignore the underlying dynamics; or lack reproducibility. Their objective in this paper is to find a balance between reality, generality, and detail.
This paper discusses six different link characteristics to consider for simulating wireless links and provides plausible models for each of them:
| Link Characteristics | How to model? |
| Error losses and corruption | Drop packets according to a per-packet, per-bit, or time-based probability |
| Delay variation | Suspend data transmission on a simulated link |
| Packet reordering | Either swap two packets in a queue, or delay one packet and let others pass |
| On-demand resource allocation | Include additional delay when a packet arrives to a queue that has been empty longer than the channel hold time |
| Bandwidth variation | Change bandwidth of a link in simulation |
| Asymmetry in bandwidth and latency | Configure up- and down-link with different values |
In addition, the paper also discusses modeling of queue management and effects of mobility. For cellular and current WLAN links, queue management can be modeled by using a Drop-Tail buffer with configurable maximum size in packets; and for satellite and future cellular and WLAN links, using RED is advised to be more effective. Mobility modeling require adequate definitions of the underlying mechanisms and depend on specific scenarios (intersystem vs intrasystem).
Comments
This paper presents a good survey of possible issues to take care of while modeling wireless links (of different types) and proposes several plausible solutions for each of the issues. The authors take no side on whether to change link layer or to change transport layer; rather, they propose a combination of the two based on suitability.