INRIA

Emmanuel.Baccelli@inria.fr http://www.emmanuelbaccelli.org/

Futurewei

+1-408-330-4586 charlie.perkins@huawei.com

Internet Internet Area I-D Internet Draft This document describes characteristics of communication between interfaces in a multi-hop ad hoc wireless network, that protocol engineers and system analysts should be aware of when designing solutions for ad hoc networks at the IP layer.

Experience gathered with ad hoc routing protocol development, deployment and operation, shows that wireless communication presents specific challenges , which Internet protocol designers should be aware of, when designing solutions for ad hoc networks at the IP layer. This document does not prescribe solutions, but instead briefly describes these challenges in hopes of increasing that awareness. As background, RFC 3819 provides an excellent reference for higher-level considerations when designing protocols for shared media. From MTU to subnet design, from security to considerations about retransmissions, RFC 3819 provides guidance and design rationale to help with many aspects of higher-level protocol design. The present document focuses more specifically on challenges in multi-hop ad hoc wireless networking. For example, in that context, even though a wireless link may experience high variability as a communications channel, such variation does not mean that the link is "broken". Many layer-2 technologies serve to reduce error rates by various means. Nevertheless, such errors as noted in this document may still become visible above layer-2 and so become relevant to the operation of higher layer protocols.

For the purposes of this document, a multi-hop ad hoc wireless network will be considered to be a collection of devices that each have at least one radio transceiver (i.e., wireless network interface), and that are moreover configured to self-organize and provide store-and-forward functionality as needed to enable communications. This document focuses on the characteristics of communications through such a network interface. Although the characteristics of packet transmission over multi-hop ad hoc wireless networks, described below, are not the typical characteristics expected by IP , it is desirable and possible to run IP over such networks, as demonstrated in certain deployments currently in operation, such as Freifunk , and Funkfeuer . These deployments use routers running IP protocols e.g., OLSR (Optimized Link State Routing ) on top of IEEE 802.11 in ad hoc mode with the same ESSID (Extended Service Set Identification) at the link layer. Multi-hop ad hoc wireless networks may also run on link layers other than IEEE 802.11, and may use routing protocols other than OLSR. The following documents provide a number of examples: AODV , OLSRv2 , TBRPF , OSPF (, and ), or DSR . Note that in contrast, devices communicating via an IEEE 802.11 access point in infrastructure mode do not form a multi-hop ad hoc wireless network, since the central role of the access point is predetermined, and devices other than the access point do not generally provide store-and-forward functionality.

In the following, we will consider several devices in a multi-hop ad hoc wireless network N. Each device will be considered only through its own wireless interface to network N. For conciseness and readability, this document uses the expressions "device A" (or simply "A") as a synonym for "the wireless interface of device A to network N". Let A and B be two devices in network N. Suppose that, when device A transmits an IP packet through its interface on network N, that packet is correctly and directly received by device B without requiring storage and/or forwarding by any other device. We will then say that B can "detect" A. Note that therefore, when B detects A, an IP packet transmitted by A will be rigorously identical to the corresponding IP packet received by B. Let S be the set of devices that detect device A through its wireless interface on network N. The following section gathers common characteristics concerning packet transmission over such networks, which were observed through experience with MANET routing protocol development (for instance, OLSR, AODV, TBRPF, DSR, and OSPF-MPR), as well as deployment and operation (e.g., Freifunk, Funkfeuer).

First, even though a device C in set S can (by definition) detect device A, there is no guarantee that C can, conversely, send IP packets directly to A. In other words, even though C can detect A (since it is a member of set S), there is no guarantee that A can detect C. Thus, multi-hop ad hoc wireless communications may be "asymmetric". Such cases are common. Second, there is no guarantee that, as a set, S is at all stable, i.e. the membership of set S may in fact change at any rate, at any time. Thus, multi-hop ad hoc wireless communications may be "time-variant". Time variation is often observed in multi-hop ad hoc wireless networks due to variability of the wireless medium, and to device mobility. Now, conversely, let V be the set of devices which A detects. Suppose that A is communicating at time t0 through its interface on network N. As a consequence of time variation and asymmetry, we observe that A: cannot assume that S = V, and cannot assume that S and/or V are unchanged at time t1 later than t0. Furthermore, transitivity is not guaranteed over multi-hop ad hoc wireless networks. Suppose that, through their respective interfaces within network N: device B and device A can detect one another (i.e. B is a member of sets S and V), and, device A and device C can also detect one another (i.e. C is a also a member of sets S and V). These assumptions do not imply that B can detect C, nor that C can detect B (through their interface on network N). Such "non-transitivity" is common on multi-hop ad hoc wireless networks. In summary: multi-hop ad hoc wireless communications can be asymmetric, non-transitive, and time-varying.

presents an abstract description of some common characteristics concerning packet transmission over multi-hop ad hoc wireless networks. This section describes practical examples, which illustrate the characteristics listed in as well as other common effects. Wireless communications are particularly subject to limitations on the distance across which they may be established. The range-limitation factor creates specific problems on multi-hop ad hoc wireless networks. Due to the lack of isolation between the transmitters, the radio ranges of several devices often partially overlap, causing communication to be non-transitive and/or asymmetric as described in . Moreover, the range of each device may depend on location and environmental factors. This is in addition to possible time variations of range and signal strength. For example it may happen that a device B detects a device A which transmits at high power, whereas B transmits at lower power. In such cases, as depicted in , B can detect A, but A cannot detect B. This exemplifies asymmetry in wireless communications as defined in .

Radio Range for Device A <~~~~~~~~~~~~~+~~~~~~~~~~~~~> | Range for Device B | <~~~~~~+~~~~~~> +--|--+ +--|--+ | A |======>| B | +-----+ +-----+

Another example, depicted in , is known as the "Hidden Terminal" problem. Even though the devices all have equal power for their radio transmissions, they cannot all detect one another. In the figure, devices A and B can detect one another, and devices A and C can also detect one another. Nevertheless, B and C cannot detect one another. When B and C simultaneously try to communicate with A, their radio signals collide. Device A may then receive incoherent noise, and may even be unable to determine the source of the noise. The hidden terminal problem is a consequence of the property of non-transitivity in multi-hop ad hoc wireless communications as described in .

Radio Range for Device B Radio Range for Device C <~~~~~~~~~~~~~+~~~~~~~~~~~~~> <~~~~~~~~~~~~~+~~~~~~~~~~~~~> | Radio Range for Device A | |<~~~~~~~~~~~~~+~~~~~~~~~~~~~>| +--+--+ +--+--+ +--+--+ | B |=======>| A |<=======| C | +-----+ +-----+ +-----+

Another situation, shown in , is known as the "Exposed Terminal" problem. In the figure, device A and device B can detect each other, and A is transmitting packets to B, thus A cannot detect device C -- but C can detect A. As shown in Figure 3, during the on-going transmission of A, device C cannot reliably communicate with device D because of interference within C's radio range due to A's transmissions. Device C is then said to be "exposed", because it is exposed to co-channel interference from A and is thereby prevented from reliably exchanging protocol messages with D -- even though these transmissions would not interfere with the reception of data sent from A destined to B.

Range for Device B Range for Device C <~~~~~~~~~~~~+~~~~~~~~~~~~> <~~~~~~~~~~+~~~~~~~~~~~> | Range for Device A | Range for Device D |<~~~~~~~~~~~~+~~~~~~~~~~~~>|<~~~~~~~~~~~~+~~~~~~~~~> +--|--+ +--|--+ +--|--+ +--|--+ | B |<======| A | | C |======>| D | +-----+ +-----+ +-----+ +-----+

Hidden and exposed terminal situations are often observed in multi-hop ad hoc wireless networks. Asymmetry issues with wireless communication may also arise for reasons other than power inequality (e.g., multipath interference). Such problems are often resolved by specific mechanisms below the IP layer; CSMA/CA, for example, requires that the physical medium be unoccupied from the point of view of both devices before starting transmission. Nevertheless, depending on the link layer technology in use and the position of the devices, such problems may affect the IP layer due to range limitation and partial overlap. Besides radio range limitations, wireless communications are affected by irregularities in the shape of the geographical area over which devices may effectively communicate (see for instance , ). For example, even omnidirectional wireless transmission is typically non-isotropic (i.e. non-circular). Signal strength often suffers frequent and significant variations, which do not have a simple dependence on distance. Instead, the dependence is a complex function of the environment including obstacles, weather conditions, interference, and other factors that change over time. Because wireless communications often encounter different terrain, path, obstructions, atmospheric conditions and other phenomena, analytical formulation of signal strength is considered intractable . The radio engineering community has developed numerous radio propagation approximations, relying on median values observed in specific environments . These irregularities cause communications on multi-hop ad hoc wireless networks to be non-transitive, asymmetric, or time-varying, as described in , and may impact protocols at the IP layer and above. There may be no indication to the IP layer when a previously established communication channel becomes unusable; "link down" triggers are often absent in multi-hop ad hoc wireless networks, since the absence of detectable radio energy (e.g., in carrier waves) may simply indicate that neighboring devices are not currently transmitting.

Many terms have been used in the past to describe the relationship of devices in a multi-hop ad hoc wireless network based on their ability to send or receive packets to/from each other. The terms used in previous sections of this document have been selected because the authors believe they are unambiguous, with respect to the goal of this document as formulated in . In this section, we exhibit some other terms that describe the same relationship between devices in multi-hop ad hoc wireless networks. In the following, let network N be, again, a multi-hop ad hoc wireless network. Let the set S be, as before, the set of devices that can directly receive packets transmitted by device A through its interface on network N. In other words, any device B belonging to S can detect packets transmitted by A. Then, due to the asymmetric nature of wireless communications: - We may say that device A "reaches" device B. In this terminology, there is no guarantee that B reaches A, even if A reaches B. - We may say that device B "hears" device A. In this terminology, there is no guarantee that A hears B, even if B hears A. - We may say that device A "has a link" to device B. In this terminology, there is no guarantee that B has a link to A, even if A has a link to B. - We may say that device B "is adjacent to" device A. In this terminology, there is no guarantee that A is adjacent to B, even if B is adjacent to A. - We may say that device B "is downstream from" device A. In this terminology, there is no guarantee that A is downstream from B, even if B is downstream from A. - We may say that device B "is a neighbor of" device A. In this terminology, there is no guarantee that A is a neighbor of B, even if B a neighbor of A. Terminology based on "neighborhood" is quite confusing for multi-hop wireless communications. For example, when B can detect A, but A cannot detect B, it is not clear whether or not B should be considered a neighbor of A; A would not necessarily be aware that B was a neighbor, as it cannot detect B. It is thus best to avoid the "neighbor" terminology, except when bidirectionality has been established. This list of alternative terminologies is given here for illustrative purposes only, and is not suggested to be complete or even representative of the breadth of terminologies that have been used in various ways to explain the properties mentioned in . Note that bidirectionality is not synonymous with symmetry. For example, the error statistics in either direction are often different for a link that is otherwise considered bidirectional.

Section 18 of RFC 3819 provides an excellent overview of security considerations at the subnetwork layer. Beyond the material there, multi-hop ad hoc wireless networking (i) is not limited to subnetwork layer operation, and (ii) makes use of wireless communications. On one hand, a detailed description of security implications of wireless communications in general is outside of the scope of this document. It is true that eavesdropping on a wireless link is much easier than for wired media (although significant progress has been made in the field of wireless monitoring of wired transmissions). As a result, traffic analysis attacks can be even more subtle and difficult to defeat in this context. Furthermore, such communications over a shared media are particularly prone to theft of service and denial of service (DoS) attacks. On the other hand, the potential multi-hop aspect of the networks we consider in this document goes beyond traditional scope of subnetwork design. In practice, unplanned relaying of network traffic (both user traffic and control traffic) happens routinely. Due to the physical nature of wireless media, Man in the Middle (MITM) attacks are facilitated, which may significantly alter network performance. This highlights the importance of the "end-to-end principle": L3 security, end-to-end, becomes a primary goal, independently of securing layer-2 and layer-1 protocols (though L2 and L1 security often help to reach this goal).

This document does not have any IANA actions.

http://www.funkfeuer.at

https://map.funkfeuer.at/wien/

http://www.freifunk.net

This document stems from discussions with the following people, in alphabetical order: Jari Arkko, Teco Boot, Brian Carpenter, Carlos Jesus Bernardos Cano, Zhen Cao, Ian Chakeres, Thomas Clausen, Robert Cragie, Christopher Dearlove, Ralph Droms, Brian Haberman, Ulrich Herberg, Paul Lambert, Kenichi Mase, Thomas Narten, Erik Nordmark, Alexandru Petrescu, Stan Ratliff, Zach Shelby, Shubhranshu Singh, Fred Templin, Dave Thaler, Mark Townsley, Ronald Velt in't, and Seung Yi.