MANET S. Ruffino Internet-Draft P. Stupar Expires: April 18, 2005 TILAB T. Clausen LIX October 18, 2004 Autoconfiguration in a MANET: connectivity scenarios and technical issues draft-ruffino-manet-autoconf-scenarios-00 Status of this Memo This document is an Internet-Draft and is subject to all provisions of section 3 of RFC 3667. By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware have been or will be disclosed, and any of which he or she become aware will be disclosed, in accordance with RFC 3668. 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. This Internet-Draft will expire on April 18, 2005. Copyright Notice Copyright (C) The Internet Society (2004). Abstract MANET interconnection with external networks enables a number of usage scenarios, but generates also a number of technical issues, mainly related with node autoconfiguration and global connectivity. This Internet Draft aims at characterizing global connectivity Ruffino, et al. Expires April 18, 2005 [Page 1] Internet-Draft MANET scenarios October 2004 scenarios and listing technical issues related to IP address autoconfiguration which are implied by such scenarios and should be addressed by a generic solution. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 5 3. Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.1 Isolated MANET . . . . . . . . . . . . . . . . . . . . . . 6 3.2 MANET connected to an external network . . . . . . . . . . 6 3.2.1 Fixed Gateways . . . . . . . . . . . . . . . . . . . . 8 3.2.2 Mobile Gateways scenario . . . . . . . . . . . . . . . 9 4. Technical issues . . . . . . . . . . . . . . . . . . . . . . . 11 4.1 Isolated MANET . . . . . . . . . . . . . . . . . . . . . . 11 4.1.1 General Aspects . . . . . . . . . . . . . . . . . . . 11 4.1.2 Stateful Autoconfiguration . . . . . . . . . . . . . . 11 4.1.3 Stateless Autoconfiguration . . . . . . . . . . . . . 13 4.2 Connected MANET . . . . . . . . . . . . . . . . . . . . . 14 4.2.1 Prefix Advertisement methods . . . . . . . . . . . . . 14 4.2.2 Prefixes assignment to the gateways . . . . . . . . . 15 4.2.3 Multiple Gateways routing and addressing issues . . . 15 4.2.4 Stateful Autoconfiguration . . . . . . . . . . . . . . 17 4.2.5 Stateless Autoconfiguration . . . . . . . . . . . . . 17 4.2.6 NAT Considerations for IPv4 . . . . . . . . . . . . . 17 4.2.7 Ingress Filtering . . . . . . . . . . . . . . . . . . 17 4.2.8 Mobile IP . . . . . . . . . . . . . . . . . . . . . . 17 4.2.9 Data Forwarding . . . . . . . . . . . . . . . . . . . 19 5. Architectural considerations . . . . . . . . . . . . . . . . . 20 6. Security Considerations . . . . . . . . . . . . . . . . . . . 21 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 22 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 24 A. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 25 Intellectual Property and Copyright Statements . . . . . . . . 26 Ruffino, et al. Expires April 18, 2005 [Page 2] Internet-Draft MANET scenarios October 2004 1. Introduction MANET were initially designed to be employed in highly dynamic and unpredicatable environments, characterized by high mobility of users and terminals. MANET are essentially autonomous, self-configuring, self-healing networks, whose mobile nodes discover other nodes and supported services in an automatic fashion. MANET routing protocols, as studied in IETF, enable two generic MANET nodes to exchange data traffic through multi-hop connections, if a 1-hop radio link between them is not available. In this way, nodes can freely move within the MANET: routing protocols dynamically react to movement and constantly discover and choose the optimal path according to a predifined metric, e.g. number of hops. If an intermediary node, which belongs to a path between a source and a destination, fails traffic is automatically re-routed through an alternative path. RFC2501 [1] gives a definition of MANET and also introduces the possibility to connect a MANET to an external network, by means of gateways: in this case the MANET acts as a "stub" network, whose nodes route traffic originating and/or terminating within the MANET only. Operators, Network and Service providers show increasing interest in this type of network, as a consequence of the wide spreading of low-cost radio technologies such as IEEE802.11a/b/g/h and the increasing customer base. Commercial MANET deployments indicate that MANETs are mainly employed as extensions of traditional infrastructure networks, to realize the so-called hybrid networks: MANET nodes exchange traffic between them using multi-hop paths and can reach outside hosts and the Internet by means of the gateways, which are equipped with two or more interfaces: a MANET interface and typically an interface to one or more non-MANET networks. An example of this are the so-called Mesh Networks, used to extend the coverage area of a public hot-spot or to realize large-scale low-cost wireless coverage in urban areas. A further interesting application and research field is represented by multi-hop cellular networks: MANETs connected to cellular WAN networks. In this case MANETs can be used to realize an extended wireless coverage in areas where "traditional" cellular network is not available. Many proposals and projects that introduce an integration between MANET and 3G+ networks exist in this area: among other [2], [3] and [4]. In both isolated and hybrid MANETs, an important technical issue is automatic configuration of network parameters on each node: IP address, default route, DNS, etc. At least one unique IP address Ruffino, et al. Expires April 18, 2005 [Page 3] Internet-Draft MANET scenarios October 2004 with local scope is needed to enable traffic exchange within a MANET. In particular, if a MANET is connected to an external network, IP address autoconfiguration rises an additional problem, known as "Global Connectivity": the IP address should be globally routable, in order for the node to be reachable both from MANET internal nodes and external Internet hosts. The intrinsic high dynamicity of MANETs rises additional issues: a MANET could split in two or more sub-MANETs, or two or more MANETs could merge in one larger MANET. In all these cases, a node could experience problems due to address conflicts, reconfiguration and reestablishment of the default route. So attention must be paid in such scenarios, expecially when multiple gateways are used within the MANET. Many solutions addressing autoconfiguration and global connectivity issues have been proposed ([5], [8], [11], [6], [7], [9]). In the authors' opinion, there is the need to define a number of MANET connectivity scenarios and, consequently, a clear problem statement, against which to validate proposed and future solutions. This document is structured in the following way: in Section 2 a glossary for commonly used terms is given; in Section 3 connectivity scenarios for a MANET are listed. In this section particular attention is given to the connection of a MANET with other external wireless networks, by means of one or more fixed (Section 3.2.1) or mobile wireless gateways (Section 3.2.2). In Section 4 IPv4 and IPv6 autoconfiguration and routing issues are derived from the previous scenarios, both for an isolated (Section 4.1) and connected (Section 4.2) MANET. Some architectural considerations are reported in Section 5. Ruffino, et al. Expires April 18, 2005 [Page 4] Internet-Draft MANET scenarios October 2004 2. Terminology Node An IPv4/IPv6 device which is a MANET element: it runs a MANET routing protocol and exchanges data with other nodes within a MANET and with hosts located within external networks. A node has at least one physical interface connecting it to the MANET. Gateway A node equipped with at least two interfaces, one of which connects it to an external network, i.e. non-MANET, and can be wired or wireless. Host An IPv4/IPv6 terminal/computer, external to the MANET. Host is defined here as only "External" to differentiate it from the nodes of the MANET. Wireless Interface (or MANET interface) The physical network interface that connects a node to the MANET. Radio Interface (or Cellular Interface) The physical network interface that may connect a gateway to an external Wireless Wide Area Network, owned and administered by an operator. Stateful autoconfiguration A mechanism of address autoconfiguration is defined stateful, when a MANET node receives its IP address by another device, acting as configuration server, which can be either within or outside the MANET. Stateless autoconfiguration A mechanism of address autoconfiguration is defined stateless when a node builds its own address without the partecipation of any external device. Default route forwarding A node transmits all data directed to external networks to one of its neighbors, elected as next-hop to the chosen gateway. Source route forwarding A node transmits all data directed to external networks explicitly indicating in each packet's header which gateway must be used to forward it. "Loose Source Routing" for IPv4 has been specified in [16]; the "Routing Header" has been specified for IPv6 in [17]. Ruffino, et al. Expires April 18, 2005 [Page 5] Internet-Draft MANET scenarios October 2004 3. Scenarios In this section, we describe the scenarios in which a MANET may typically operate and where different address autoconfiguration issues arise. 3.1 Isolated MANET An isolated MANET is a network that is autonomously set-up between wireless mobile nodes localized in the same geographical area. Nodes activate Layer 2 radio links, by which they can exchange traffic with their neighbors, and employ an ad-hoc routing protocol, which enables multi-hop data forwarding through intermediate nodes. Routing protocol constantly discovers routes between nodes, in a proactive ([12], [14]) or reactive fashion ([13], [15]): this enables each node to route traffic to all other nodes within the MANET also during movements. In this type of MANET there is no connection to an external network: all traffic is generated by MANET nodes and addressed to MANET nodes. Typical applications of this scenario are temporary networks, that must be set-up in areas where neither wireless coverage nor infrastructure exist. Examples can be emergency networks used for disaster recovery, battlefield applications, electronic surveillance. Other examples can be found in occasional work meetings, where networks are formed to enable file sharing between co-workers. 3.2 MANET connected to an external network In this scenario a MANET is connected to an external network by means of one or more gateways (Figure 1). A generic MANET node, running a MANET routing protocol, can exchange data traffic with every other node through multi-hop paths and communicate with hosts located in the external network, routing its uplink traffic towards a gateway. This, in turn, will receive return traffic from the host and will route it to the source node. Ruffino, et al. Expires April 18, 2005 [Page 6] Internet-Draft MANET scenarios October 2004 H1 | +---------------+ | Internet |** +---------------+ * * * * * * * GW1** * GW3 | +--GW2-------+ | | | ---N1--------+ | / \ | N4 \ N2 N3-----/ Figure 1: MANET interconnected to an external network Gateways play a critical role here. If the number of nodes in the MANET increases, gateways can become bottlenecks, as they route an increasing and possibily huge amount of traffic. This also depends on the available bandwidth on the uplink interface. Moreover, gateways can be equipped with a number of additional features. For example, they could participate to the external routing protocol, in order to announce internal routes to external routers and hosts, possibly performing some kind of aggregation. They can act as enforcement points for security purposes: they can control access to external networks and, following a common practice, they can enforce Ingress Filtering on MANET generated traffic. Finally they can also provide services like DNS to MANET nodes. This scenario can be expanded, depending on the characteristics of the network interface connecting gateways to the external network: it can be either wired or wireless, which can, in turn, be of a different type with respect to the MANET interface. In the first case Gateways are fixed, while in the second case they can also be mobile, as the other MANET nodes. Moreover, a MANET can have only one gateway (fixed or mobile) or can have multiple gateways (fixed or mobile). Other than guaranteeing a higher degree of reliability and fault tolerance to the entire MANET, the presence of multiple gateways permits load balancing among the gateways themselves. This can be very useful especially when the external network is a low-throughput cellular WAN, such as GPRS/EDGE, in order to not overload a single gateway with traffic potentially generated by many nodes at the same time. Single traffic flows of multiple nodes or many flows of a single node can be routed through Ruffino, et al. Expires April 18, 2005 [Page 7] Internet-Draft MANET scenarios October 2004 different gateways, consequently suggesting an improvement of the overall performances of the MANET. Gateways can also be equipped with additional resources in order to grant better fault tolerance to the entire MANET: additional energy resources, more processing power, more volatile and non-volatile memory. This is especially true in case of fixed gateways, that can be directly powered and directly operated. It is clearly more difficult to enhance mobile gateways, due to their limited energy resources. The following sections detail usage scenarios for fixed and mobile gateways. 3.2.1 Fixed Gateways In this scenario, gateways are deployed in predefined positions planned by the network operator. Each gateway is connected to the external network by means of a wired or wireless interface. Mesh networks and networks used for environmental surveillance can be categorized under this scenario. o Mesh Networks: these are probably the most widespread ad-hoc networks. In a Mesh Network, user terminals (nodes) exchange traffic between them directly through a layer-2 radio link and using other nodes or fixed wireless Access Points as intermediate relays. A Mesh Network is typically connected to an external infrastructure network by means of fixed wired Access Points, which act as gateways and typically connect the Mesh to an external infrastructure network. Mesh Networks can be further categorized, depending on if the mesh is realized only between the wireless Access Points or also between all the nodes, which, in this case, run a routing protocol. In fact, in some deployments, intermediate access points are equipped with two wireless interfaces: the first interface forms the mesh with other peer access points, participating in an ad-hoc routing protocol, the second interface provides local connectivity to nodes, which cannot set-up a network themselves, as they don't run any routing protocol. Applications of this networks are Internet public access (browsing, email etc.) by mobile users from outdoor areas, wireless coverage of corporate building to give employees access to shared data and commonly used services (email, Intranet browswing). These solutions can bring to savings on cabling costs. Ruffino, et al. Expires April 18, 2005 [Page 8] Internet-Draft MANET scenarios October 2004 o Surveillance networks: several wireless nodes endowed with sensors of varios kinds are spread over high enviromental risk areas (e.g. fires). They communicate through multi-hop connections and run a routing protocol. When an emergency situation arises, data collected by sensors are transmitted from the collecting nodes upwards one or more gateways (which can have both a wired or wireless interface) and conveyed to a manned monitoring station. Topologies of this kind of network are typically static, as the nodes are installed in fixed positions within the monitored areas. Moreover, these networks are characterized by multiple constant low-throughput data flows going from the sensors to the gateways. 3.2.2 Mobile Gateways scenario In this scenario, the gateway's radio interface, connecting the MANET to the external network, can be a cellular WAN interface (GSM, GPRS, EDGE, UMTS), a broadband wireless MAN (WMAN) interface (e.g. 802.16x, 802.20) or a WLAN interface (802.11a/b/g/h/j). In each of these cases, gateways can forward uplink traffic outside the MANET only if located within the transmission/reception range of one or more Base Stations or Access points. Gateways can therefore not only freely move within the coverage area, but they can also move outside this area: in such case, the gateway can't forward uplink traffic destined for external hosts anymore, nor downlink traffic destined for internal nodes. The primary benefit of coverage extension is that local communication between two nodes of the MANET are preformed without using any cellular radio resource, e.g. radio channels. Another benefit is the possibility to grant network access also to those terminals that are not equipped with a cellular radio interface (e.g. access sharing). The implication of this business model on security and accounting aspects are out of the scope of this draft. A more advanced scenario can be realized when most of the nodes are also equipped with two heterogeneous interfaces. In this case gateways can be "occasional": they can be nodes that, after setting up the connection towards the external network, whenever located within its coverage area, can start forwarding other nodes' outbound packets. In this kind of scenario, gateways can be "special" nodes endowed with additional features, but they can also be ordinary MANET nodes, such as mobile phones and PDAs. In this last case, gateways are characterized by low computational power and limited energy resources. Although the MANET can again exploit benefits given by multiple gateways, additional issues arise: in fact, gateways are not under operators control anymore. It's possible that the owner of the gateway decides abruplty to turn his terminal off or to tear down the connection towards the cellular network, in order to save battery Ruffino, et al. Expires April 18, 2005 [Page 9] Internet-Draft MANET scenarios October 2004 life. Thus, the number and the position of gateways are higly dynamic and this can cause frequent re-routing of uplink data flows. In this situation, autoconfiguration operations performed by nodes and gateways assume critical importance, as they have not only to be performed by a node on joining the MANET, but also constantly repeated, in order to repair broken routes to the gateway. Ruffino, et al. Expires April 18, 2005 [Page 10] Internet-Draft MANET scenarios October 2004 4. Technical issues The scenarios introduced in the previous sections, raise some technical issues that must be addressed both to guarantee the normal functionalities of a MANET and to assure an accettable level of performance to the applications running on each single node. Among those issues there is particullary the one of the nodes' IPv4/IPv6 address configuration. The purpose of this section is not to propose of a solution to the address configuration issue, but to characterize and describe the technical aspects, which are related to such issue, affect the scenarios descripted above and lack of a standard solution. 4.1 Isolated MANET In an isolated MANET scenario, the nodes' addressing issue has been already considered by many proposals which furnish a set of solutions. See [21] for a complete and exaustive comparison of such proposals. However, some of these proposals do not cope with all the critical aspects that affect the autoconfiguration issue in an isolated MANET. The purpose of this section is to describe the main critical aspects related to the addressing in this scenario. 4.1.1 General Aspects As an isolated MANET is characterized by flat addressing: a node must only be endowed with an address, unique within the MANET itself. There are no limitations on the prefix of such address, both when using IPv4 and when using IPv6. Therefore, the problem can be reduced to the configuration of an unique IP (IPv4 or IPv6) address, without any constraint on the chosen prefix. Such configuration should be made in an efficient and automatic way. In the following, stateful and stateless approaches are considered separately. 4.1.2 Stateful Autoconfiguration In an isolated MANET, the role of the configuration server can be only assigned within the MANET itself and can be performed by any node of the network. It must be always guaranteed that within a MANET there is a node acting as configuration server. This may not be always true, as a MANET can experience a partition during its Ruffino, et al. Expires April 18, 2005 [Page 11] Internet-Draft MANET scenarios October 2004 existence. If indeed there is only one node providing stateful configuration within a MANET and such a MANET gets partitioned, one of the newly-born networks will be without any connection to the configuration server. Several solutions have been proposed to counter this issue: o Dynamic Election: the role of the configuration server is dynamically assigned to one only MANET node. o Server Redundancy: the role of configuration server is distributed among more than one terminal of the MANET; it can also be distributed among all the MANET nodes. When an isolated MANET is endowed with a stateful address mechanism, there is the problem of detecting, by means of a Service Discovery mechanism, which of the MANET nodes acts as a configuration server. This is true both when the configuration service is assigned to a single node and when this role is distributed among several nodes. Moreover, when the stateful configuration service is not distributed among all the nodes, a MANET which has been created by a previous partition can be without any configuration server: the problem of the configuration server election within such MANET therefore arises. Partitioning is not the only event creating some problems with respect to addressing. It can't be excluded that two MANETS merge, creating a new MANET, as such kind of networks are characterized by the mobility of their elements. Moreover it can't be excluded that two nodes, located in two different MANETs, own the same address (the unicity of an address is indeed guaranteed only within the MANET the node is connected to). If these two MANETs merge with each other, there would be a newly-born MANET in which two nodes have configured the same address. Such an event surely compromises the normal functions of these two nodes, but also of the whole network. It is therefore necessary that a node verifies the unicity of its address after a merging. Such requirement implies the existence of a merging-detection mechanism, or that every node keeps on verifying the univocity of its address. It is worth mentioning that when the stateful configuration is executed by means of more than one server, it is possible that within a MANET, two of such servers assign the same address to two nodes: the problem of the DAD execution therefore arises, otherwise it is necessary that a coordination mechanism exists among servers, which assures that assigned addresses belong to disjoint addressing space. Ruffino, et al. Expires April 18, 2005 [Page 12] Internet-Draft MANET scenarios October 2004 4.1.3 Stateless Autoconfiguration If the configuration mechanism is stateless, there are several issues, depending on the used Internet protocol. A stateless configuration mechanism has been standardized both for IPv4 and for IPv6. Such mechanisms are based upon the assumption that all the nodes of the network are located on the same link: this assumption is not valid for MANETs. Moreover, such solutions don't guarantee that the configured address is unique, and therefore a DAD mechanism is still needed. We assume here that a global prefix is not available in the MANET. o In case of IPv4, the standard stateless configuration mechanism is ZEROCONF [19]: through its use, a node configures itself a link-local address derived from the prefix 169.254.0.0/16. o In the case of IPv6 the standard stateless configuration is defined IPv6 stateless autoconfiguration [18] and assures the automatic configuration of a link-local address on network interfaces by using the well known prefix fe80::/16. The suffix of such address is the EUI-64 identifier, which is derived from IEEE 802 identifier of network interfaces through which the nodes are connected to the network. The uniqueness of such addresses is not guaranteed if the network is made of nodes having heterogeneous interfaces, in particular when one or more of them are connected to the network by means of interfaces that have no IEEE 802 identifier (e.g. Bluetooth). Such nodes must be endowed with a mechanism that lets them generate a suffix which will be used to set up the link-local address through stateless configuration. It is, however, not guaranteed that the generated address is unique and that a DAD mechanism is therefore still required. The issues related to the use of link-local addresses within a MANET, listed in [5] , can be summarized by asserting that IPv6 prevents the nodes from communicating with such type of addresses along multi-hop paths. o In ZEROCONF, the DAD mechanism is based upon ARP messages exchange, which is used by the nodes to defend their address and to detect the presence of a duplicate address. In IPv6 the same mechanism is based upon exchange of Neighbor Solicitation and Neighbor Advertisement messages. In both cases (IPv4 and IPv6), the DAD mechanism has been defined to be executed within LANs, which are networks whose elements are on the same link. As cited before, in a MANET, such an assumption is not valid: it is therefore necessary to introduce a standard DAD mechanism which is able to detect that two nodes use the same address even when such nodes are not directly connected on the same link. Ruffino, et al. Expires April 18, 2005 [Page 13] Internet-Draft MANET scenarios October 2004 o It can't be excluded that two nodes located into two different MANETS own the same address also when the configuration mechanism is stateless and not only in the case of a stateful approach: if these two MANET merge, there will be two nodes having the same address in the new MANET. It is therefore necessary that MANET nodes verify the uniqueness of their address not only when they enter the MANET, but also after a merging. Such requirement implies the necessity of a merging-detection mechanism or the necessity of a continuos DAD execution, as stated in section Section 4.1.2. For example, [19] and [18] define the continuos defense of nodes' addresses. Incidentally, partitioning do not imply any issue, as addresses are trivially unique before and after the MANET gets partitioned. 4.2 Connected MANET If the MANET is connected to an external IP network, e.g. the Internet, the address configuration must let the MANET nodes to communicate with hosts located in the external network. In turn, this implies the configuration of an IPv4 or IPv6 global address: the network interface partecipating to the MANET routing protocol must be configured with at least one globally routable address, which can be public or private (only for IPv4). In case of public addressing, such an address must be derived from a prefix which must be globally valid, while in the case of private IPv4 addressing there must be a NAT agent somewhere. We assume here that prefixes are managed and announced by gateways in the MANET. Several issues related to global IP address configuration can be characterized. Some of these are similar to those described in section Section 4.1, however, there are some scenarios that create further aspects that must be considered. The most complex scenarios are particularly the ones characterized by the presence of multiple gateways and multiple prefixes. In such situations a MANET indeed experiences high dynamicity, as described in section Section 3.2. 4.2.1 Prefix Advertisement methods It is firstly necessary considering the ways through which prefixes are announced within a MANET. Two general approaches exist: prefix information is inserted into routing protocol messages or inserted into other protocol messages. In the first case it is necessary an opportune extension of routing protocol: the solutions could not be general enough, as the global connectivity is tied to every single routing protocol. Using a routing protocol can optimize the diffusion, the choice and the management of the prefixes that must be used: a deeper description of such aspect is given in Section 4.2.3. Viceversa, if prefix advertisement is independent from the routing Ruffino, et al. Expires April 18, 2005 [Page 14] Internet-Draft MANET scenarios October 2004 protocol, it is necessary the definition of new mechanisms or the modifications to existing ones. IPv6 Neighbor Discovery, for example, must be adapted to the multi-hop nature of the MANET paths. 4.2.2 Prefixes assignment to the gateways For this purpose, gateways must own the prefixes which they announce within the MANET. It is possible to assume that they are manually configured on gateways, but this approach is problematic if it is applied to dynamic scenarios such as those described in Section 3.2.2. In this scenario, gateways can appear or disappear within a MANET in an unpredictable way. Gateways should therefore automatically receive the global prefixes associated to the MANET they are connected to, e.g. through a prefix delegation mechanism [23], that, moreover, lets external routers install automatically the proper routes towards the MANET. 4.2.3 Multiple Gateways routing and addressing issues The presence of multiple active gateways can improve the overall robustness of the MANET, as described in Section 3.2. In fact, gateway redudancy guarantees a higher fault tolerance: for example, if a gateway fails, nodes, which are using such gateway, can dynamically change their own default gateway and continue to communicate with external hosts. Multiple gateways can improve global MANET performance if nodes are enabled to concurrently route traffic through more than one gateway. If all the MANET nodes used the same gateway as default gateway, leaving the others as pure back-up, load balancing among different gateways would not be exploited. Moreover, if uplink traffic is supposed to be sent towards different gateways and if also downlink traffic is supposed to be received from different gateways, the choice of the prefix used by a node to configure its address assumes critical importance. In fact, IP source address used for uplink traffic determines the gateway to which the return traffic will be sent. It is desiderable that traffic follows the same path uplink and downlink within the MANET: this is because the uplink path towards the default gateway is guaranteed to be optimal by routing the protocol. However, in case of multiple gateways / multiple prefixes in the MANET, the following can arise (as depicted in Figure 2): a node N1 elects a default gateway GW1, receives prefixes form it, configures its global address, begins a data session with an external host H1. Both uplink and downlink traffic are delivered through GW1. Then the node moves away from its position and find itself near another Ruffino, et al. Expires April 18, 2005 [Page 15] Internet-Draft MANET scenarios October 2004 gateway, GW2, which has a better routing metric than the previous gateway, which is now many hops away (namely, N4 - N3 - N2). The node elects this as default gateway, but keeps its old global address, in order not to disrupt the session. Downlink traffic follows in this case a non-optimal path within the MANET, which can be potentially very long: the bandwidth available to the user (and to other MANET nodes) may descrease dramatically. It is therefore desiderable that the return traffic flows towards the default gateway of the source node. ---------->H1 / | / +---------------+ / +-| Internet |---+ / | +---------------+ | | | | | GW1++ +GW2 | | \ / v | \N2----N3----N4/ N1 -----------H1<------------ / | \ / +---------------+ \ / +-| Internet |---+ \ / | +---------------+ | | | | | | | GW1++ +GW2+ | \-----\ \ / | | \ \N2----N3----N4+ | | \--------------\ \----N1 ------^ Figure 2: Multiple gateways non-optimal routing In this scenario, a critical situation arises when one or more gateways are no longer reachable from MANET nodes (e.g. because MANET gets partitioned or gateways fail or simply shut-off). In this case, nodes, whose global addresses are associated to such gateways , are no more globally reachable and should configure a new address. Mobile IP can handle this address change. See Section 4.2.8 for considerations on this approach. Ruffino, et al. Expires April 18, 2005 [Page 16] Internet-Draft MANET scenarios October 2004 4.2.4 Stateful Autoconfiguration Differently from what happens in an isolated MANET, in hybrid MANETs the configuration server can also be located in the external network. This option requires more attention and a deeper analysis, particulary in the scenarios of Section 3.2.2, where gateways can activate and deactivate abruptly. The major issue here comes from the fact that a node may not reach the external server, thus experimenting high delays and session disruption. 4.2.5 Stateless Autoconfiguration If the connected MANET uses a stateless autoconfiguration mechanism, same considerations as in Section 4.1.3 arise. In this case, the node must be able to configure a global address: this must be derived from a globally valid prefix, which is also used to install routes towards the MANET within the external network. Issues related to discovery of such prefixes and DAD have been described in Section 4.2.1 and Section 4.1.3, respectively. 4.2.6 NAT Considerations for IPv4 If any NAT mechanism is deployed, the problem of the prefix diffusion, which is described in Section 4.2.1 and Section 4.2.2, is avoided as the node in the MANET can choose a random address and use it to partecipate to the routing protocol. However, the nodes of a MANET whose gateways use NAT mechanism to achieve global connectivity, as described in [24], experience a session break if they change the gateway used to send uplink traffic. 4.2.7 Ingress Filtering Ingress filtering is a security mechanism often implemented on sites border routers, which could also be deployed on gateways of a connected MANET. It consists in blocking all the outbound packets whose source address does not belong to a prefix advertised within the MANET. This means that in a MANET with multiple gateways, which advertise multiple prefixes, nodes that have received a prefix from a gateway must also use the same gateway to route uplink traffic. In such a MANET, a node which changes its default gateway, must also change its address, in order to be able to continue to transmit traffic towards external networks. 4.2.8 Mobile IP This section highligths some issues raised by the use of Mobile IP in MANETs. Ruffino, et al. Expires April 18, 2005 [Page 17] Internet-Draft MANET scenarios October 2004 In many of the scenarios described above, particularly in those of Section 3.2, a MANET node may have to change its IP address. Reasons for this can be the following: o a node moves to a separate MANET, where different global prefixes are advertised; o the gateway, that is advertising the prefix used to configure a node's address, fails: in this case downlink traffic directed to such prefix cannot be delivered to the MANET anymore; o ingress filtering (see Section 4.2.7 ) is deployed in the MANET to improve security; o as seen in Section 4.2.3 a node could change its address in order to use always an optimal path both for uplink and downlink traffic, thus acheiving better performances. In order to handle this address changes, Mobile IP ([20]) can be deployed on MANET nodes. Mobile IP is commonly used to avoid data session disruption after a mobile node changes its IP address as a consequence of a change of point of attachment to the Internet. A MANET with multiple gateways can be considered as an overlapping set of different access networks, if gateways advertise multiple global prefixes. The use of standard Mobile IP (either for IPv4 or IPv6) in this scenario can imply several issues, described in the following. o In the scenario described in Section 4.2.3, where multiple egress gateways can concurrently be active, a node could change its address frequently. This brings to the "Binding Storm" problem: if many nodes in the MANET change their addresses at the same time, a very high number of Binding Update messages (or Registration Request, if Mobile IPv4 is used) will be generated, therby overloading the MANET and gateways in particular. o If Mobile IPv4 is used, some nodes with special functions, such as Foreign Agent or DHCP server, must be located within the MANET in order to assign a valid address to the nodes. If the Foreign Agent approach is used, two more mechanisms are needed: a discovery mechanism for Foreign Agents and a mechanism which enables multi-hop transmission of Advertisement and Registration Request/Response messages within the MANET. If the DHCP approach is used, same consideration hold as those described in Section 4.2.4. Moreover, if robustness is needed and one employs redundant agents, synchronization issues may arise. Ruffino, et al. Expires April 18, 2005 [Page 18] Internet-Draft MANET scenarios October 2004 o If Mobile IPv6 is used, the major issue is related to Movement Detection. In fact, Movement Detection is realized by means of Router Advertisements: if these messages are not present in the MANET, i.e. prefixes are advertised through MANET routing messages, standard Mobile IPv6 procedure will not work. Moreover, Router Advertisements were not designed to be multi-hop, so same considerations as in Section 4.1.3 and Section 4.2.1 apply. 4.2.9 Data Forwarding There can be two methods which a node can use to forward data traffic towards external networks : default route and source routing. Both methods, applied in the reference scenarios described above, have pros and cons. If the default route method is employed, intermediate nodes of a path from a source node to a gateway may not have elected the same default gateway as the source node. This means that data can flow towards a different egress gateway: if ingress filtering is deployed, communication is not possible. If source routing is used, data can flow through a gateway explicitly chosen by source node. Moreover, if a default route is used, load balancing in the MANET could be problematic as, again, nodes can't control the path data follow through the gateway. Ruffino, et al. Expires April 18, 2005 [Page 19] Internet-Draft MANET scenarios October 2004 5. Architectural considerations There are many different proposal that solve many of the issues described above. As already mentioned, the objective of this draft is not to give a survey of state of the art solutions, nor to address problems that such solutions pose. It is, however, needed to clarify what architectural choices are more effective, in order to design a comprehensive solution for both autoconfiguration and global connectivity problems. Two important aspects are the following. o Layering : this means integration between different protocols (autoconfiguration, mobility, routing). It can bring the definition of optimized mechanims, exploiting information passed from one protocol to another (e.g. to perform movement detection). From a different point of view, it is however desiderable to keep different functionalities separated from each other, in order to get a modular approach. o Nodes with specific functionalities: in some scenarios one can deploy some special nodes, enabled with mechanisms such as DHCP, NAT, Foreign Agent, gateway, that have, for example, a fixed connection to external network and are equipped with more energetic resources than "normal" nodes. In this case these special nodes perform all "operational" tasks and normal nodes are only source or destination for data traffic and are only requested to run a MANET routing protocol. In this case, if a special node fails, unpredictable results are possible, that can bring to the failure of the entire MANET. Moreover, in a very large MANET (i.e. big cardinality), it can be difficult to mantain an acceptable level of performance with only one active gateway, which can become a bottleneck for outbound traffic. Ruffino, et al. Expires April 18, 2005 [Page 20] Internet-Draft MANET scenarios October 2004 6. Security Considerations This document raises no security issue. Ruffino, et al. Expires April 18, 2005 [Page 21] Internet-Draft MANET scenarios October 2004 7. IANA Considerations This document has no actions for IANA. 8 References [1] Corson, S. and J. Macker, "Mobile ad hoc networking (MANET): Routing protocol performance issues and evaluation considerations", RFC 2501, January 1999. [2] Siebert, M., "On Ad Hoc Networks in the 4G Integration Process", Med-Hoc 2004 , June 2004. [3] "Ambient Networks", http://www.ambient-networks.org . [4] "World Wireless Research Forum", http://www.wireless-world-research.org . [5] Wakikawa, R., Malinen, J., Perkins, C., Nilsson, A. and A. Tuominen, "Global connectivity for IPv6 Mobile Ad Hoc Networks", I-D draft-wakikawa-manet-globalv6-03.txt, October 2003. [6] Cha, H., Park, J. and H. Kim, "Extended Support for Global Connectivity for IPv6 Mobile Ad Hoc Networks", October 2003. [7] Jeong, J., Park, J., Kim, H. and D. Kim, "Ad Hoc IP Address Autoconfiguration", I-D draft-jeong-adhoc-ip-addr-autoconf-02.txt, February 2004. [8] Perkins, C., Malinen, J., Wakikawa, R. and E. Belding-Royer, "IP Address Autoconfiguration for Ad Hoc Networks", I-D draft-perkins-manet-autoconf-01.txt, November 2001. [9] Singh, S., Kim, JH., Choi, YG., Kang, KL. and YS. Roh, "Mobile multi-gateway support for IPv6 mobile ad hoc networks", I-D draft-singh-manet-mmg-00.txt, June 2004. [10] Paakkonen, P., Rantonen, M. and J. Latvakoski, "IPv6 addressing in a heterogeneous MANET-network", I-D draft-paakkonen-addressing-htr-manet-00.txt, December 2003. [11] Jelger, C., Noel, T. and A. Frey, "Gateway and address autoconfiguration for IPv6 adhoc networks", I-D draft-jelger-manet-gateway-autoconf-v6-02.txt, April 2004. [12] Clausen, T. and P. Jacquet, "Optimized link state routing protocol", RFC 3626, October 2003. Ruffino, et al. Expires April 18, 2005 [Page 22] Internet-Draft MANET scenarios October 2004 [13] Perkins, C., Belding-Royer, E. and S. Das, "Ad hoc On-Demand Distance Vector (AODV) Routing", RFC 3561, July 2003. [14] Ogier, R., Templin, F. and M. Lewis, "Topology Dissemination Based on Reverse-Path Forwarding (TBRPF)", RFC 3684, February 2004. [15] Johnson, D., Maltz, D. and Y. Hu, "The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks (DSR)", I-D draft-ietf-manet-dsr-10.txt, July 2004. [16] Postel, J., "Internet Protocol", STD 5, RFC 791, September 1981. [17] Deering, S. and R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", RFC 2460, December 1998. [18] Thomson, S. and T. Narten, "IPv6 Stateless Address Autoconfiguration", RFC 2462, December 1998. [19] Aboba, B., "Dynamic Configuration of Link-Local IPv4 Addresses", draft-ietf-zeroconf-ipv4-linklocal-17 (work in progress), July 2004. [20] Johnson, D., Perkins, C. and J. Arkko, "Mobility Support in IPv6", RFC 3775, June 2004. [21] Sun, Y. and E. Belding-Royer, "A study of dynamic addressing techniques in mobile ad hod networks", I-D Wireless communication and mobile computing, May 2004. [22] Narten, T., Nordmark, E. and W. Simpson, "Neighbor Discovery for IP Version 6 (IPv6)", RFC 2461, December 1998. [23] Troan, O. and R. Droms, "IPv6 Prefix Options for Dynamic Host Configuration Protocol (DHCP) version 6", RFC 3633, December 2003. [24] Engelstad, P., T—nnesen, A., Hafslund, A. and G. Egeland, "Internet Connectivity for Multi-Homed Proactive Ad Hoc Networks", First IEEE International Conference on Sensor and Ad hoc Communications and Networks , October 2004. Ruffino, et al. Expires April 18, 2005 [Page 23] Internet-Draft MANET scenarios October 2004 Authors' Addresses Simone Ruffino Telecom Italia LAB Via G.Reiss Romoli 274 Torino 10148 Italy Phone: +39 011 228 7566 EMail: simone.ruffino@telecomitalia.it Patrick Stupar Telecom Italia LAB Via G.Reiss Romoli 274 Torino 10148 Italy Phone: +39 011 228 5727 EMail: patrick.stupar@telecomitalia.it Thomas Heide Clausen Laboratoire d'informatique Ecole Polytechnique Palaiseau Cedex 91128 France Phone: +33 1 6933 2867 EMail: thomas.clausen@polytechnique.fr Ruffino, et al. Expires April 18, 2005 [Page 24] Internet-Draft MANET scenarios October 2004 Appendix A. Acknowledgments The authors would like to thank Ivano Guardini for his valuable comments. Ruffino, et al. Expires April 18, 2005 [Page 25] Internet-Draft MANET scenarios October 2004 Intellectual Property Statement The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org. Disclaimer of Validity This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Copyright Statement Copyright (C) The Internet Society (2004). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. Acknowledgment Funding for the RFC Editor function is currently provided by the Internet Society. Ruffino, et al. Expires April 18, 2005 [Page 26]