IRTF ANS Research Group L. Yang Internet Draft S. Conner Document: draft-irtf-yang-ans-scenarios-00.txt X. Guo Expires: April 2004 M. Hazra J. Zhu Intel October 2003 Common Wireless Ad Hoc Network Usage Scenarios Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026 [1]. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract This document examines some common wireless ad hoc network usage scenarios that have the potential of being deployed as medium or large scale networks. The scenarios include both pure ad hoc networks without any infrastructure support and the hybrid networks with some infrastructure in place. This document is motivated by the effort in the IRTF ANS (Ad hoc Network Scalability) research group. The aim of this document is to provide some common network scenarios that are rooted in the real world deployment to facilitate meaningful investigation and research of scalability in ad hoc networks. Yang, et al. Expires - April 2004 [Page 1] Common Ad Hoc Network Usage Scenarios October 2003 Table of Contents 1. Introduction..................................................2 1.1 Terminology Overview......................................2 1.2 Motivations...............................................4 1.3 Document Structure........................................6 2. Enterprise Networks...........................................6 2.1 Environmental Parameters..................................6 2.2 Independent Parameters....................................8 2.3 Primary Metrics...........................................8 3. Community Networks............................................9 3.1 Environmental Parameters..................................9 3.2 Independent Parameters...................................11 3.3 Primary Metrics..........................................11 4. Hot Spot Networks............................................11 4.1 Environmental Parameters.................................12 4.2 Independent Parameters...................................13 4.3 Primary Metrics..........................................14 5. Home/Apartment Networks......................................14 5.1 Environmental Parameters.................................14 5.2 Independent Parameters...................................15 5.3 Primary Metrics..........................................16 6. Summary......................................................16 Security Considerations.........................................17 References......................................................17 Acknowledgments.................................................18 Author's Addresses..............................................18 1. Introduction One of the primary focuses of the ANS (Ad hoc Network Scalability) research group is to evaluate the scalability of ad hoc routing protocols developed by the IETF MANET community. Since it is fairly expensive and time consuming for most researchers to deploy medium to large scale networks in the real world for the purpose of research and evaluation, simulation is the predominant method in ad hoc network research, esp. for scalability study. Therefore, it is very important to understand the characteristics of real world network scenarios so that the simulation is done with meaningful and realistic assumptions. We also hope that by introducing and using a common set of network scenarios, the IRTF ANS group and the larger ad hoc research community can share and compare their research results more directly and easily. 1.1 Terminology Overview To some people, the terminology of Mobile Ad Hoc Networks (MANET), or simply, Ad Hoc Networks, is used synonymously with Mobile Mesh Networks, and Mobile Multi-hop Wireless Networks. Nevertheless, these terminologies may mean slightly different things for others. So for the sake of discussion in this document, we would like to define the relevant terminologies more precisely. According to RFC 2501 ([1]), "A MANET consists of mobile platforms (e.g., a router with multiple hosts and wireless communications devices)--herein simply referred to as 'nodes' -- which are free to move about arbitrarily". "A MANET is an autonomous system of mobile nodes. The system may operate in isolation, or may have gateways to and interface with a fixed network. In the latter operational mode, Yang, et al. Expires - April 2004 [Page 2] Common Ad Hoc Network Usage Scenarios October 2003 it is typically envisioned to operate as a 'stub' network connecting to a fixed internetwork. Stub networks carry traffic originating at and/or destined for internal nodes, but do not permit exogenous traffic to 'transit' through the stub network." While the definition above remains valid and accurate, we do want to point out some signs of shifting trend as to how MANET technology can be used in the real world deployment. MANET research was initially motivated by the military and defense sector where the main appeal of an ad hoc network is its peer-to-peer communication capability in a dynamic and mobile environment without any form of infrastructure support. However, the recent wide spread success of IEEE 802.11 WLAN technology in the consumer, enterprise and service provider markets demonstrates that Internet or Intranet connectivity remains the primary application driver and so infrastructure support remains a key part of most wireless mobile networks. Therefore, it is necessary to make the distinction of a pure ad hoc network (or simply, an ad hoc network) from a hybrid network, as defined below. A hybrid network is typically consisted of two distinct categories of nodes. One category is called the infrastructure nodes, (e.g., Access Points) which are typically put in place with the primary purpose of aggregating and transporting the traffic for the client nodes in the network. The other category of nodes in a hybrid network is the client nodes, which represent the application users of the network. In another word, a hybrid network is consisted of two hierarchical levels -- infrastructure support and clients. A pure ad hoc network, on the other hand, consists of a flat hierarchy of only client nodes without any infrastructure nodes. Most of the networks deployed today are hybrid networks instead of pure ad hoc networks. It is worth noting that both the infrastructure nodes and the client nodes can be wireless and mobile nodes, and so multi-hop mesh technology can be potentially used to form both a wireless infrastructure mesh, and a client mesh. We adopt the terminology of "mesh network" to refer to the multi-hop network where each node can relay traffic on behalf of other nodes. "Mesh" can be formed at different hierarchy in the networks. As we will point out later in each scenario, some usage scenario may form mesh in both hierarchical layers while some may form mesh only in one but not both. On the other hand, we use "ad hoc network" only to refer to the flat mesh networks consisting of only client nodes but no infrastructure nodes. Mobility has been a salient feature in most of the MANET research, partly due to the fact that an ad hoc military network is indeed consisting of highly mobile nodes in the battle field. Mobility is still a very important feature and sometime a primary motivation for most wireless networks deployed in the commercial world. However, the kinds of typical client nodes (e.g., laptop, PDA, consumer devices at home, etc.) and the applications prevalent in the commercial sector (email, web surfing, file sharing, etc.) suggest a very different usage model and hence a very different mobility model, as it will become apparent in the following sections as we discuss each of the usage scenarios. For example, people typically sit down with their laptops before checking their email; and people tend to stand still or walk only slowly when using their PDA, unless they are inside a car or a train. So mobility does exist but maybe not as prevalent as previously assumed in a military environment. In the following sections we will discuss the mobility pattern for each Yang, et al. Expires - April 2004 [Page 3] Common Ad Hoc Network Usage Scenarios October 2003 usage scenario individually. Overall we assume that the nodes can be potentially mobile but they may also remain fixed in some scenarios. It is also worth pointing out that mobility is not the only factor contributing to the dynamic nature in a wireless network. For example, an indoor wireless network consisting of only fixed (non- mobile) wireless nodes still demonstrates very dynamic link characteristics over time, typically due to the movement of objects and people around the network nodes in the environment. Therefore, a wireless network with fixed nodes can still face the similar challenge (e.g., dynamic links between neighboring nodes) as a network with highly mobile nodes. Some research attention has been given to mobility modeling so that the models can represent the real world mobility more realistically. However it remains an open question whether the mobility of the other objects (including people) in the environment results in the same kind of dynamic link characteristics as the mobility of nodes themselves would. It remains to be seen whether the current mobility models commonly used by the Ad Hoc research community can model such dynamic environment realistically. If not, the question becomes how to model not just the node mobility but also the dynamic environment for wireless networks. 1.2 Motivations [2] provides a working definition of scalability for a method of an ad hoc network. According to [2], the scalability of a method in an ad hoc network is a measure of its ability to maintain efficiency in a set of metrics, with respect to a given environment, as some independent parameter of the network increases to very large values. Therefore, the scalability of a method is evaluated with respect to an [INDEPENDENT PARAMETER, PRIMARY METRICS, ENVIRONMENTAL PARAMETERS] triple, where INDEPENDENT PARAMETERS are parameters that can be freely varied; the PRIMARY METRICS are performance quantities that are observed in the network; and the ENVIRONMENTAL PARAMETERS are parameters that define the operational conditions of the network. For example, researchers may evaluate a particular routing protocol's scalability by evaluating the performance curves of end-to-end throughput and delay versus increasing number of nodes in a certain network environment. Some useful independent parameters listed in [1] include: - Number of nodes - Node density - Traffic load - Mobility Primary metrics of the system are the dependent variables that are observed as the network is scaled with respect to an independent parameter. In general, multiple primary metrics are arranged in a SCALABILITY VECTOR and scalability of a certain method is evaluated with respect to this vector. Some common primary metrics include: - (1/throughput) of the network - Delay in the network - Battery power required at the network nodes - Memory/storage required at the network nodes - Processing power required at the network nodes - 1/(network lifetime) Yang, et al. Expires - April 2004 [Page 4] Common Ad Hoc Network Usage Scenarios October 2003 Environmental parameters are the most diverse one among the triple. Many environmental parameters may affect scalability, including categories such as: - Network characteristics - Node characteristics - Traffic pattern - Routing layer being used - MAC layer being used - PHY layer being used Each category above can include many different individual variables. Consider the collection of all the possible combinations of these environmental parameters as a multi-dimensional "environmental space", and each combination of the environmental parameters becomes only a single data point in this space. It becomes apparent that this is a huge high-dimensional space and it is practically impossible to evaluate any give method for all the possible data points within this space. Researchers in the ad hoc network community typically evaluate a given method only for a few data points in this huge environmental space. But the choices of these data points are somewhat arbitrary, and it becomes difficult to leverage or compare with other researchers' results because the choices of these data points usually vary from one researcher to another. Practically speaking, not every data point in this space is equally interesting. Ad hoc network is an important technology for wireless networks and the markets for today's wireless networks include home, small offices, large enterprises, campus networks, hot spots, community networks, military networks, sensor networks etc. Most of these networks have the potential to be deployed at medium to large scales, with the exception of maybe single-family houses and small offices. If we examine each of these categories carefully, it becomes obvious that each has some distinct characteristics in terms of network topology, traffic pattern, mobility pattern, nodes etc. Each of the categories can be represented more or less as a dense cluster of data points instead of random data points in the huge environmental space. We argue that these dense clusters represent some of the most interesting data points in this huge environmental space, because they represent some of the most common ad hoc network usage scenarios deployed in the real world. The objective of this document is to identify a subset of common ad hoc network usage scenarios in this environmental space with description of the environmental characteristics, esp. the network characteristics, the node characteristics and the traffic pattern. Some of the other categories of the environmental parameters, e.g., the routing protocol, MAC and PHY layer used, are more technology dependent and so in general remain unspecified in the descriptions provided here. However, 802.11 WLAN technology has been the major driving force in today's wireless markets and so many of the scenarios described here are typically deployed with 802.11 infrastructure support in the real world. Pure ad hoc networks (i.e., without any infrastructure support) are more suitable for some usage scenarios (like military networks and sensor networks) but not others. Nevertheless, we believe there is real value in combining the multi-hop networking technology into the infrastructure-based usage scenarios to create hybrid networks. Therefore, most of the scenarios described in this document belong to hybrid networks rather than pure ad hoc networks. Yang, et al. Expires - April 2004 [Page 5] Common Ad Hoc Network Usage Scenarios October 2003 In addition to describing some of the environmental characteristics, this document also tries to identify the most relevant independent parameters and the most important scalability metrics when the networks in each category scale upward toward medium to large scale networks. 1.3 Document Structure Rest of this document descries some of the common wireless network usage scenarios, including both ad hoc and hybrid networks. Each of the following sections is devoted to one common network usage scenario. The scenarios included in this version of the documents are: - Enterprise networks - Hot spot networks - Community networks - Home/Apartment networks. It is our intention to include more usage scenarios in the future version of the document, including sensor networks, military networks, vehicle networks on the highway, etc. At the end of this document, we also point out some of the initial findings from studying the usage scenarios described in this document. 2. Enterprise Networks As the name suggests, enterprise wireless networks deal with the deployment and management of a wireless LAN in the enterprise. Wireless LAN technology can be used to link corporate buildings in diverse locations. Use of wireless technology in the enterprise network has several advantages. - Ease of installation. Using wireless LAN technology, IT professionals can extend wired LANS and easily provide connectivity to any location in the enterprise. Dynamic moves, additions and changes are easy to implement. - Low cost. The need to install expensive wiring is reduced. IT professionals installing networked computers can save money by using a cost-effective wireless network infrastructure solution. - Mobility. Wireless Enterprise networks keep workers productive wherever business takes them by providing the mobile workers secure, high-speed access to critical corporate network resources even when they are away from their desks. 2.1 Environmental Parameters TodayÆs enterprise wireless networks are typically deployed with IEEE 802.11 infrastructure mode, and the "ad hoc" mode as defined by 802.11 is seldom used in such networks. Enterprise networks generally have very large infrastructure, with 100s of APs and 1000s of wireless clients, covering most of the indoor space on the enterprise campus. All client devices associate with an AP and all APs are connected to the wired backbone. However, we do see the potential of utilizing multi-hop mesh technology among the APs and even among the fixed client nodes in the near future, so tomorrow's enterprise networks likely become true hybrid networks. Yang, et al. Expires - April 2004 [Page 6] Common Ad Hoc Network Usage Scenarios October 2003 The enterprise typically has very good control over all the nodes participating in the enterprise networks, therefore, it represents one of the most controlled usage scenarios with most stringent performance requirement. 2.1.1 Node Characteristics The nodes in the enterprise wireless network can be categorized as: - Access Points (APs): In an enterprise network today, each AP is connected to the wired backbone. Client devices associate with an AP, which then acts as the clientÆs gateway to the enterprise network. Currently most of the APs are connected into the enterprise network via wired medium. In the future, it is foreseeable that a wireless multi-hop mesh network can be formed among the APs via wireless medium. In the AP mesh networks, all APs need not be connected to the gateway directly. Such a wireless backhaul can help further reduce the need for wiring in the enterprise buildings. - Client Devices: The client devices can be further broken down into two sub-categories. o Mobile Clients. This includes laptops with wireless capabilities and wireless Personal Digital Assistants (PDAs). These devices have a limited battery life available, especially when mobile and not connected to a power source. When static these nodes may be utilized in forwarding traffic to other nodes. When in motion, however, these devices can switch their association from one AP to another and are not very reliable in providing uninterrupted forwarding service. Also the power (battery life) limitation of these devices makes it infeasible to have the nodes on for a long time to act as relay nodes. o Fixed clients. This includes desktops equipped with wireless NIC cards. The wireless enabled desktop client devices have higher processing power and higher memory and storage capabilities than the mobile clients. They are walled powered devices and can be potentially always on. Being fixed in wall-powered locations and having comparatively better resources than the mobile clients, and being owned by the enterprise (instead of the personnel using them), they have the potential of being used as relay nodes to forward traffic for other client nodes in a multi-hop fashion. Security will be an important issue that needs to be addressed when implementing multi-hop networking using these devices. 2.1.2 Network Characteristics In the enterprise deployment of a wireless LAN today, physical site surveys are done to place the APs. IT professionals and network planners take several factors into account while doing the site survey. The site survey helps in selecting the number of APs, location of the APs and each APÆs wireless settings. Once the site survey is done, APs are distributed more-or-less uniformly in each floor of a building and the density of the APs is also uniform. Hence a regular topology, e.g. grid or hexagon, is commonly used to place the infrastructure nodes for todayÆs enterprise networks. As offices/cubicles and conference rooms are more or less uniformly laid out through the buildings, the fixed clients nodes tend to Yang, et al. Expires - April 2004 [Page 7] Common Ad Hoc Network Usage Scenarios October 2003 conform to uniform or regular topologies. On the other hand, the mobile client nodes tend to be either randomly distributed throughout the building (as people move around with their laptops) or densely cramped into a few small locations (e.g., inside the conference rooms). In future enterprise networks, a mesh topology may exist between the APs; i.e. a wireless backhaul may exist between the APs configured into a mesh topology. A mesh network can also exist between clients of the enterprise network, utilizing the more reliable fixed desktops as relay for the other clients. Or a hybrid ad hoc network can exist where an AP equipped with multiple radios can use one radio to be part of the AP mesh and the other to be part of the client mesh. Enterprise networks typically are deployed in the indoors environment, and the wireless nodes are surrounded with many objects (including poles, walls, furniture) and walking people that can cause highly dynamic RF characteristics for the wireless links, even if the nodes stay static. Attenuation of the RF signal at the crowded office environment may be more severe than the outdoors and so the radio transmission range may also be shorter than that in the outdoors. So the PHY layer model needs to take into account both the node mobility and environmental dynamic. 2.1.3 Traffic Pattern The mobile worker typically needs access to the Internet, his personal e-mail or corporate databases. Hence most traffic is data centric, dealing with web-surfing, e-mail and file sharing applications. However need for voice (for VoIP applications) and video (for net-meeting like applications) is growing, in order to allow workers to keep in touch with their colleagues while in motion. Most of the traffic is directed between the client nodes and the APs and between the APs and the wired gateway. Hence the performance requirements for the enterprise WLAN is more stringent than in other WLAN scenarios; i.e. there is need to provide high-bandwidth access to the mobile users. In addition, as there is need to protect company secrets, need to secure the wireless communication is very high in this scenario. 2.2 Independent Parameters As mentioned earlier, when deploying the WLAN in the enterprise, IT personnel and network planners take into account several factors. Of these, the independent parameters, that affect the network design (how many APs to add, where to place the APs, how to configure the APsÆ settings) include: o Coverage Area: The area over which coverage is to be provided. o Node density: The expected number of wireless clients in the coverage area and the expected number of mobile clients in the coverage area. o Load: Expected load to be offered by each client to the network. 2.3 Primary Metrics The most important metrics in the enterprise WLAN are throughput per client and delay encountered by a client in the network. The aim is Yang, et al. Expires - April 2004 [Page 8] Common Ad Hoc Network Usage Scenarios October 2003 to provide the mobile user seamless roaming capabilities, by offering him a throughput and delay experience similar to what he can get at his desk. 3. Community Networks The widespread availability of WLAN technology has led to a large number of grassroots groups around the world to use this technology to share Internet connectivity across neighborhoods, parks, college campuses, and city blocks. WLAN technology can be used to provide "last-mile" broadband connectivity to homes in areas that do not have cable modern or DSL service as well as to mobile users outdoors. Unlike wireless networks that are configured and maintained by service providers (e.g., hotspot networks, as described in Section 4), community networks are typically deployed by individuals or non- profit organizations or donated by companies to provide free or very low-cost wireless network access. It is difficult to define a "typical" community network since these networks tend to be deployed in a very ad hoc manner, wherever resources happen to be available (i.e., very limited planning). Likewise, they tend to consist of non-uniform topologies since nodes are often deployed wherever convenient rooftops, lamp posts, or hilltops happen to be available and owned by individuals/ organizations that are willing to donate real estate to the network deployment effort. 3.1 Environmental Parameters Community networks tend to cover large geographical areas including both indoors and outdoors spaces, like across neighborhood, parks, city blocks, etc. A combination of IEEE 802.11 (WLAN), IEEE 802.16 (WMAN) and other technologies (some of them proprietary) maybe used to provide different functional layers needed to cover the intended area. The very nature of the community networks being owned/formed by a group of volunteers with limited resources makes it easier to leverage the multi-hop mesh technology in these networks to share the network connectivity, extend the coverage and lower the cost for the community as whole. 3.1.1 Node Characteristics The devices that comprise a community network tend to be heterogeneous due to different functional layers of typical community networks. Most community networks deployed today consist of Access Point (AP) devices deployed at fixed locations where wired Internet access happens to be available and mobile client devices that connect to an AP for Internet access. In addition, some community networks leverage multi-hop wireless backhaul to extend range and coverage from wired network locations. Summary of node types: - Long-Range Wireless Router Nodes: devices that are deployed in strategic locations to provide long-range wireless backhaul to wired Internet access. These devices may have multiple radios and directional antennas for long-range line-of-sight connectivity. These devices are deployed in fixed locations with available line- power. - Access Point Nodes: devices deployed at fixed locations to provide wireless access to fixed and mobile clients. Unlike long-range wireless router nodes, these typically do not need to be deployed at strategic locations with line-of-sight to distant long-range Yang, et al. Expires - April 2004 [Page 9] Common Ad Hoc Network Usage Scenarios October 2003 wireless router nodes. They may connect directly to a nearby long-range wireless router node, connect using dense mesh topologies (usually with omni-directional antennas), or have a local wired Internet connection. These devices typically have line power, and may be co-located with a long-range wireless router node. - Fixed and Mobile Client Nodes: end-user devices that connect to access point nodes for Internet access. These devices may connect directly to an AP or use ad hoc routing to leverage other mobile client nodes to connect to an AP. The devices may be in fixed locations with line-powered, but are often mobile with battery power. 3.1.2 Network Characteristics Community networks tend to be deployed in a very ad hoc manner, wherever resources such as wired Internet access and antenna/radio mounting locations happen to be available. Community networks are usually deployed by volunteers in a best-effort manner, unlike enterprise and service provider networks (hot spots) that are designed to provide as uniform coverage as possible. This often leads to very non-uniform topologies. As described in the previous section, community networks tend to consist of multiple functional topology layers. The primary purpose of a community network is usually to provide connectivity to the Internet, although there is usually not a single wired backbone. Nodes tend to be initially deployed in sparse topologies to provide initial connectivity in convenient locations. They tend to be relatively small scale when initially deployed, but grow in scale over time. Summary of topology elements: - Long-range wireless backhaul of routers, most likely using directional antennas and multiple radios. Some of these devices are connected directly to wired internet connections, while multi-hop wireless protocols are used to connect the remaining devices via wireless backhaul. - Clusters of APs connecting to long-range wireless routers, either with directional or omni-directional antennas. - Dense mesh of APs (shorter range than the wireless backhaul routers), some of which have either a wireless link to a long-range wireless router or their own wired network connection. These could have one or multi radios and could use either directional or omni antennas. - End-point client device connectivity. Each of the APs could act as a local community hotspot, providing wireless access to clients (e.g., within a home, business, park, etc). 3.1.3 Traffic Pattern The primary purpose of most community networks is to provide connectivity to the Internet. End-devices connect primarily to destinations on the Internet (outside of the wireless network itself), e.g. for web surfing and email applications. A fraction of the communication may consist of file sharing and multi-player game traffic that requires peer-to-peer connectivity between end-devices in the wireless network. Because community networks typically provide wireless Internet access for free or very low cost, performance requirements are typically less stringent than in service provider networks. In general, a community network should provide Yang, et al. Expires - April 2004 [Page 10] Common Ad Hoc Network Usage Scenarios October 2003 higher throughput per-user than is available from alternative low- cost network connection options (e.g. 56k modem). 3.2 Independent Parameters When community networks are deployed using multi-layer topologies, there is opportunity to divide scalability analysis into largely independent pieces. For example, the maximum throughput that can be provided by each long-range wireless router when the backhaul is connected with directional antennas can be studied primarily independently from the throughput per client that can be achieved surrounding a given backhaul router depending on the number and topology of APs and clients. Parameters that can be independently configured when simulating community networks include: - Long-Range Wireless Router Topology - Access Point Density surrounding a Long-Range Wireless Router or wired network connection - Client Density - Traffic Load 3.3 Primary Metrics The most important metrics in real-world community networks are typically cost and coverage. Because end-users typically do not pay for service, the absolute throughput per user is usually not as critical as in service provider hotspot networks. However, when analyzing different community network configurations, the following metrics should be considered: - Coverage - Throughput per client - Latency 4. Hot Spot Networks A major appeal of WLANs is that network access providers can quickly deploy them in areas with a high volume of users of network traffic and where the traditional wireless carriers have not been able to keep up with public demand for broadband connections. These high traffic areas are referred as Hot Spots. Several thousand hot spots have already been identified throughout the world and the number is expected to increase significantly over the next several years. Dozens of start-up hot spot operators have installed Wi-Fi access points at hotels, airports, restaurants, bookstores, schools, theaters, convention centers, health clubs, and other public venues. Major wireless carriers are now entering the WLAN connectivity business leveraging their nationwide coverage and extensive customer base, e.g. T-Mobile offers 802.11b-based access for more than half of the hot spots in the United States, with tens of thousands of mobile users accessing the Internet from its WLANs. The hot spot networks deployed today are typically based on 802.11 infrastructure mode. As we will discuss below, we believe it is reasonable to expect the hot spot operators may leverage mesh technology to build up the wireless backhaul among APs for larger coverage (meaning the traffic may be relayed by several APs over wireless links before reaching its final destination at the gateway); Yang, et al. Expires - April 2004 [Page 11] Common Ad Hoc Network Usage Scenarios October 2003 but it is unlikely for the hot spot operators rely on the client nodes to act as traffic relay for other clients. 4.1 Environmental Parameters The geographic area covered by the hot spot networks can vary a lot, depending on the intended location and the business model of the service provider. A small hot spot covering a coffee shop or a restaurant may only require one or two APs, while a hot spot operated throughout an international airport may require 50-100 APs. Larger size of hot spots may share many similar characteristics with the community networks. For example, the mixture of technologies being used (802.11, 802.16 etc.), the functional layers (long range wireless backhaul routers, APs, and clients) deployed to cover a larger geographic area may be very similar to what we describe for community networks in Section 3. On the other hand, the very different business model of hot spots (owned and operated by service providers to make money) also makes it very different from the community networks in some aspects. For example, while it is relatively easy to convince the volunteers of a community network to allow other people using their APs as relay nodes to get on the Internet; it will be a big challenge to ask the paying customers to do the same on their devices. 4.1.1 Node Characteristics There are mainly two kinds of nodes in a hot spot û client nodes and Access Point. The former includes mobile terminal such as laptops, hand held devices, and so on. Access Point is deployed at fixed locations to provide wireless access to mobile clients. If the geographic area is large, long range wireless backhaul router may be used with multiple radios and/or directional antennas, just as in the community networks. As power outlets are typically under short supply in the hot spots, the client nodes are typically operated with limited battery power. Hand held devices are also very resource constraint in terms of the memory and processing power. As customers walk in and out of the hot spots, the client nodes join and leave the networks very frequently. Although the environment in a hot spot is highly dynamic, the mobility of most client nodes is typically very low because most people tend to sit down while working on their laptops and the walking speed while working with hand held also tend to be slow. 4.1.2 Network Characteristics The key characteristic of the hot spots deployed today is heterogeneous connectivity. Usually, the network size could range from small to very large, where the number of Access Points could range from few (e.g. < 5) to a large value (e.g. 50). The typical topology for a hot spot today is a star topology. All client nodes are associated with one of the APs and all APs are connected to the gateway on the wired backbone because network connectivity remains the major application for hot spot customers. However, in the future, it is possible especially for a large scale hot spot with multiple cells that APs are connected with each other wirelessly and form a wireless backhaul. In a hot spot, client nodes are randomly distributed and the density typically is not uniform or even regular throughout the geographic Yang, et al. Expires - April 2004 [Page 12] Common Ad Hoc Network Usage Scenarios October 2003 area. This leads to the co-existence of links with different signal quality; therefore, one of the main challenges in a hot spot is heterogeneous link connections. For 802.11 WLAN, a direct consequence of the heterogeneous connections is a multi-rate environment, because IEEE 802.11 provides different data rates for different link quality. As presented in many previous research works, current 802.11 MAC protocol guarantees packet level fairness, where all links get the equal share of the total throughput no matter what data rate it uses. As a result, the whole system performance can be very sensitive to the distribution of nodes, since a node with a slow link can drag down the whole system throughput for other nodes. Comparing to community networks, the service provider has more control over the coverage and density of the infrastructure nodes deployed in a given facility. The service providers may still deploy the APs in a very non-regular way, for example, so that better coverage is achieved in the sitting areas than along the corridors in the airport to match the expected usage pattern by the end users. Given the dynamic and resource constraint natures of the client nodes in a hot spot, and the business model most hot spots are operating under, it is very difficult to make a client device work as a relaying node to forward traffic for others willingly and reliably. Security will also be another important obstacle for using client devices as relays in a hot spot. Therefore we expect the hot spot operators continue relying on their own infrastructure to deliver the customers' traffic. On the other hand, as geographic area of the hot spot and the number of client devices using the hot spot grow larger and larger, the operators have to put in more and more APs to support the growing demand. While wireline backbone is difficult and expensive to deploy for a large area, the wireless backhaul becomes an attractive alternative. Therefore, we expect multi-hop capable wireless backhaul become an enabling technology component to deploy large scale hot spot networks. 4.1.3 Traffic Pattern Internet-based applications will be the main applications for the hot spots, such as email and web, etc. Most traffic is data, and so it is not very delay sensitive. Local peer-to-peer traffic will be very small comparing to the out-going/in-coming traffic through gateway. Therefore, the traffic pattern for the hot spots is very much star- like, and the down-link (in-coming) traffic dominates due to the nature of the main applications (email, web, etc.). As we have pointed out in the previous section, it may be very difficult to implement multi-hop networking among the client nodes in a hot spot due to many issues, such as security and limited resources provided by laptop. Thus, the average number of hops between a client node and an infrastructure node is one, but mesh may still be used among the APs to reach the final gateway to the Internet. 4.2 Independent Parameters Yang, et al. Expires - April 2004 [Page 13] Common Ad Hoc Network Usage Scenarios October 2003 The main factor driving the scaling dimension is the total number of clients in a hot spot. The out-going bandwidth is limited (say T1 line) thus the per-client throughput decreases as the number of clients increases. The performance of contention-based MAC protocol in IEEE 802.11 is likely to drop significantly as the number of clients increases because of the increased contentions. Therefore, as the number of clients increase, the number of APs in the hot spot tends to increase as well. 4.3 Primary Metrics The main metrics in a hotspot is throughput per client. 5. Home/Apartment Networks The initial driving force for wireless (e.g. IEEE 802.11 networks) at home/apartment is its convenience at extending broadband access to devices (primarily laptops) anywhere in the home. But the future home networks will include more wireless-capable devices such as the consumer electronic devices and can potentially become a hybrid networks with local ad-hoc communications amongst these devices. 5.1 Environmental Parameters Today's single-family home network is typically a small scale network with a few client devices connecting to one or two 802.11 APs. However, as higher throughput radio technology (with shorter range) becomes available, it is foreseeable to form a multi-hop mesh networks among the client devices at home. That can be achieved by turning every consumer devices into a wireless mesh relay node, or by developing specialized home-use wireless relay nodes to form a low cost infrastructure mesh inside a home. Medium to large scale networks can be formed in a sizable apartment complex or dormitory building on college campus. This may be the result of a well planned activity when a single administrative entity (i.e., the apartment owner) deployed the infrastructure throughout the building; or it may be the result of a bunch of uncoordinated small networks set up by neighboring apartment tenants inside their own apartment units. In either case, coordination among the client nodes across the boundary of apartment units to form a mesh may be a challenge because these client nodes belong to different tenants. Forming a mesh among APs may be feasible if these APs are owned by a single entity. 5.1.1 Node Characteristics Nodes in the future home network may include computing devices (e.g. desktops, laptops, handhelds), multimedia devices (e.g. video recorder, HDTV, home theatre), appliances and possibly low cost consumer grade wireless routers (special purpose wireless relay nodes). These devices have a wide range of computing, communication and resource capability. Most of the home appliances (DVD, TV, etc.) are plugged to the wall. So will the wireless routers. Power consumption is not the primary concern. However, home appliances may employ low cost commodity wireless component. Hence the processing capacity (computing, storage) may not be as powerful as a laptop. Therefore, protocols Yang, et al. Expires - April 2004 [Page 14] Common Ad Hoc Network Usage Scenarios October 2003 or mechanisms designed for home appliance network should be scalable to processing capacity. Most of the client devices remain fixed in locations (e.g., consumer devices) while some may be mobile (laptop, handheld, etc.). 5.1.2 Network Characteristics Co-channel interference (CCI) is not an issue in a single home environment where neighbors are far enough such that interference from them can be ignored. But CCI becomes significant in apartment complex environment where homes share walls. Frequency planning is necessary. However, frequency selection amongst apartment units may lack the level of coordination found in enterprise or hotspot environments. Hence, frequency planning for apartment complex can not take advantage of a single administrative domain, and may need to assume an un-cooperated environment. Frequency agility may also be required to avoid interference from non-networking devices using the same (unlicensed) frequency band. Such devices include microwave oven, cordless phone, baby monitor, etc. A home network may need to support multimedia service that requires QoS guarantee. Hence the routing and reservation scheme should be accurate. However in a home environment, even a short distance between two devices may experience multiple wall attenuation and complex multi-path propagation. So it is very difficult to accurately measure the link capacity between deices. Hence the QoS scheme should be able to tolerate such inaccuracy. In addition, it should explore multi-path routing for better QoS guarantee. 5.1.3 Traffic Pattern A home network may need to support a large variety of traffic patterns. Low data rate such as sensory data from appliance, medium data rate such as web browsing or high data rate such as HDTV streams. While traffic going in/out of the home networks to connect to the outside world is definitely a significant part of the total traffic expected in a home network, we also expect to see significant peer-to-peer traffic within the local home network. For example, media files (music or even movie clips) may be fed from the desktop PC (hided somewhere in the corner of the den/study) to the media players (DVD, TV, etc.) in the family room via the local mesh networks formed inside the house or apartment unit. Therefore, home networks may distinguish itself from many other scenarios (enterprise, community, hot spots, etc.) because of its unique traffic pattern. Traffic variation due to mobility is not expected to be significant because many of the client nodes remain fixed. Similar to the indoors enterprise networks, the RF links may still be highly dynamic due to the multi-path attenuation of the signals and the mobility of people (and pets) inside the home. 5.2 Independent Parameters Load -- the most significant parameter will be the number of high- throughput applications that need to be supported simultaneously. Yang, et al. Expires - April 2004 [Page 15] Common Ad Hoc Network Usage Scenarios October 2003 5.3 Primary Metrics From the perspective of serving multimedia applications, the primary metrics include throughput and delay. Both the average and the distribution will be evaluated, as little jitter in both metrics are desirable. From the perspective of robust connectivity (e.g. to long-running multimedia applications), the primary metrics may also include route resilience to path/node failure. From the perspective of serving heterogeneous traffic classes, the primary metrics may also include the ability to isolate each traffic class and provide performance predictability. 6. Summary Even though we've only examined a few of the common usage scenarios in which ad hoc networking technology may be applicable, we can see that some of the common assumptions used by many researchers in the ad hoc research community are not always valid in the real world. One common assumption people use for most ad hoc networking research (but admittedly less so in the last two years), is that the ad hoc network consists of only homogeneous nodes in a flat hierarchy and each and every node can act as both a client and a relay for other nodes in the network. We can see that is not always true for the scenarios we describe here. Most of the scenarios deploy some kind of infrastructure hierarchy and so we should evaluate the protocol or method in question for both the pure ad hoc networks and the hybrid networks to ensure real world success. Another common assumption used by many of the researchers is that all the nodes in the networks are mobile and can move about arbitrarily. As pointed out by some of the most recent research in mobility study, most of the mobility models commonly used today don't match well with the mobility patterns in the real world. As we point out earlier, node mobility is not the only factor contributing to the RF dynamics in the wireless networks. Mobility of objects and people in the environment contribute probably just as much as the mobility of the nodes themselves. Little research is done so far to evaluate and model the dynamic effect on the RF links and routing method. Picking random pairs of source and destination nodes in the network constitutes a common traffic pattern assumed by many research papers' simulation. However, as we examine the scenarios in this document, most of the traffic patterns are star-like traffic between client nodes in the network and the gateway nodes that can connect to the Internet or Intranet. As we discuss in the introduction section, the environmental space in which we evaluate any ad hoc protocol or method can be a huge multi- dimensional space with infinite evaluation data points. We believe a set of well defined and understood usage scenarios can go a long way to help the research community focus our collective research and evaluation effort around a few important cluster data points in the vast environmental space. The usage scenarios we describe in this document are only a few examples that we intend to use as starting Yang, et al. Expires - April 2004 [Page 16] Common Ad Hoc Network Usage Scenarios October 2003 points for the discussion in the IRTF ANS group and the larger ad hoc research community. We just pointed out a few findings that came out of this study that might help us better tune our assumptions and models in the future simulations or even test-bed study. Security Considerations This document only describes some of usage scenarios either exist today or potentially might emerge in the near future. Cecurity consideration is one of the aspects we look at for each scenario. Since this document does not define or recommend any networking protocol or specific technology and hence there is no directly applicable security consideration that is relevant here. References [1] S. Corson, J. Macker, ôMobile Ad hoc Networking (MANET): Routing Protocol Performance Issues and Evaluation Considerations", RFC 2501, January 1999. [2] O. Arpacioglu, T. Small, Z. J. Haas, ôNotes on Scalability of Wireless Ad Hoc Networks", work in progress, August 2003, . 1 Bradner, S., "The Internet Standards Process -- Revision 3", BCP 9, RFC 2026, October 1996. Yang, et al. Expires - April 2004 [Page 17] Acknowledgments Author's Addresses Lily Yang Intel JF3-206 2111 NE 25th Ave., Hillsboro, OR 97124 Phone: (503)264-8813 Email: lily.l.yang@intel.com Steve Conner Intel JF3-206 2111 NE 25th Ave., Hillsboro, OR 97124 Phone: (503)264-8036 Email: w.steven.conner@intel.com Xingang Guo Intel JF3-206 2111 NE 25th Ave., Hillsboro, OR 97124 Phone: (503)264-9944 Email: xingang.guo@intel.com Mousumi Hazra Intel JF3-206 2111 NE 25th Ave., Hillsboro, OR 97124 Phone: (503)712-7853 Email: mousumi.m.hazra@intel.com Jing Zhu Intel JF3-206 2111 NE 25th Ave., Hillsboro, OR 97124 Phone: (503)712-7235 Email: jing.a.zhu@intel.com ang Expires - April 2004 [Page 1]