Locator ID Separation for Mobility Management in the New Generation Network - PDF

Please download to get full document.

View again

of 13
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report

Pets & Animals


Views: 5 | Pages: 13

Extension: PDF | Download: 0

Related documents
Locator ID Separation for Mobility Management in the New Generation Network Ved P. Kafle and Masugi Inoue National Institute of Information and Communications Technology (NICT) Tokyo, Japan {kafle,
Locator ID Separation for Mobility Management in the New Generation Network Ved P. Kafle and Masugi Inoue National Institute of Information and Communications Technology (NICT) Tokyo, Japan {kafle, Abstract Given that the majority of communications devices are mobile terminals, efficient mobility support should be a key feature in the new generation network or future Internet. The current Internet does not have native mobility support. Although variants of Mobile IP protocols have been developed to address this problem, these protocols cause signaling overhead, create a single point of failure, and lack smooth handover capabilities and interoperability between IPv4 and IPv6. Recently, locator ID separation has been considered as a promising approach to better mobility support, while also improving security and routing scalability. In this paper, we present the mobility-related functions of the recently proposed locator ID separation-based network architectures. We also outline their limitations and list some possible extensions that will be needed if they are to be deployed in the new generation network. 1 Introduction The Internet does not support mobility natively because IP addresses have an overloaded semantic of both host identifiers (IDs) and locators [1]. Namely, an IP address is used in network layer protocols as a locator to find the destination host in the network topology and forward packets toward their destination. The same IP address is also used in the transport and upper layer protocols as the host ID to identify the host or sessions associated with the host. These layers use IP addresses in session IDs by binding the socket application program interface (API) with the IP address. When the host moves from one subnet to another and connects through a new attachment point, it acquires a new IP address, while invalidating the previous IP address. This terminates the session identified by the previous IP address. Recently, variants of Mobile IP protocols [13, 14] have been developed to resolve the mobility limitations of Internet architecture. However, these protocols cause signaling overhead, create a single point of failure, and lack smooth handover capabilities and interoperability between IPv4 and IPv6. The introduction of the locator ID separation concept, also known as ID/locator split (i.e., using separate namespaces for host IDs and locators) approach to network architecture simplifies mobility support functions. The Internet Engineering Task Force (IETF) and other segments of the Internet community have recently been discussing this approach as a concept that could not only aid mobility but also contribute to multihoming, routing scalability, and security [6, 7, 10]. Related work in ID/locator split-based mobility architectures includes the Host Identity Protocol (HIP) [10, 11], Location Independent Network Architecture (LIN6) [12], and Locator/ID Separation Protocol (LISP) [7]. However, since the introduction of the locator ID separation approach will significantly change the Internet architecture, this approach is more suitable for the new generation network or future Internet, which would be based on a clean-slate design. Some future network projects, such as the AKARI Project [4] and the 4WARD Project [5]), are designing network architectures on the basis of this concept. Journal of Wireless Mobile Networks, Ubiquitous Computing, and Dependable Applications, volume: 1, number: 2/3, pp This is an invited paper. 3 This paper presents a comparative study of the mobility functions available in related work. It identifies those functions that have not been included in architectures such as HIP and LISP, which have been progressing under the auspices of the IETF, but are covered in the AKARI Project [4]. The Heterogeneity Inclusion and Mobility Adaptation through Locator ID Separation (HIMALIS) architecture [3] of the new generation network is being developed as part of the AKARI Project. The HIMALIS architecture provides mobility functions for handover optimization and supporting heterogeneous network layer protocols. Its mobility functions are performed by the identity layer, a new layer inserted between the network and transport layers, which supports mobility across different network layer protocols. This paper is organized as follows. Section 2 discusses related studies, the lessons learned from them, and their limitations. Section 3 describes the mobility functions of ID/locator split architectures, with a focus on the HIMALIS architecture. Section 4 lists additional issues such as security and interoperability with other functions (e.g., multihoming, routing) of the ID/locator split-based mobility architecture. Section 5 concludes the paper. 2 Related Work 2.1 Mobile IP Protocols Recognizing the limitations of the original Internet architecture in supporting mobility, a number of Mobile IP protocols [13, 14] have recently been developed and standardized in the IETF. These protocols enable a mobile host to possess two types of IP addresses: a home address and a care-of address. The home address, which is configured from the home network IP address prefix, is the persistent address. The mobile host can continue to use the home address even when it moves to a foreign network. The mobile host gets a care-of address in the foreign network and registers the binding between the home address and the care-of address in the home agent located in the home network. The home address is anchored at the home agent. That is, data packets sent at the home address do reach the home agent when the mobile host is not located in the home network. The home agent forwards these packets to the mobile host s care-of address after encapsulating them with an additional IP header. As the packets arrive at the mobile host, the Mobile IP functions installed in Layer 3.5 de-capsulate the packets and forward them to the upper layer (i.e., transport layer). Mobile IP protocols such as [13, 14] are host-based mobility management protocols. In other words, the protocol functions are implemented in mobile hosts that detect movements and carry out location update signaling with the home agents. There are additional protocols to optimize the handover operation, either by localizing mobility signaling messages or by creating local tunnels between access routers to reduce packet loss during handovers. Hierarchical Mobile IP [15] confines the flow of signaling messages within the visited network domain and Fast Mobile IP [16] establishes a tunnel between access routers to forward packets destined for the mobile host from the previous access router to the new access router during handovers. Similarly, certain network-based mobility management protocols have also been standardized. The network mobility (NEMO) [17] and Proxy Mobile IP [18] protocols fall in this category, where the mobility management functions are executed not by the individual host but by network nodes such as the mobile routers or access routers. By observing the Mobile IP related protocols development activities in the IETF, we have learned that the new generation network should have both host- and network-based mobility support functions. It should also possess functions for supporting network mobility and reducing handover delay as well as packet losses, while minimizing signaling traffic. For this purpose, we give an overview of locator ID separation-based mobility management protocols being developed in the IETF [7, 11] and the AKARI project [2, 3]. The mobility management scheme should also be independent of the network-layer protocols so that the same scheme can be applied to both IPv4 and IPv6 as well as to any future network 4 protocols. 2.2 ID/Locator Split-Based Mobility Protocols HIP [10, 11], LINA [12], LISP [7], and HIMALIS [3] fall into the category of ID/locator split-based mobility protocols because they use different sets of values for host IDs and locators. HIP [11] uses public keys (and their hash values) as host IDs and IP addresses as locators. A new layer, called the identity layer, inserted between the transport and network layers of the host protocol stack performs the host ID-to-locator mapping functions. HIP extends the Domain Name System (DNS) records to store host IDs as well. A host acquires its peer host s ID and locator by sending a domain name lookup request to a DNS server. While communicating with the peer host, both the source and destination hosts IDs appear in the identity header and locators in the network header of data packets. Although HIP is a good step in developing a locator ID separation-based mobility scheme, it is still in its infancy and lacks several functions. It has no support for smooth handover. Its session initiation process is computationally heavy, making it inappropriate for small, resource-limited devices. It uses locators in some signaling messages, thus necessitating the re-establishment of session contexts in the event of switching locators. This requirement is counterproductive to fast handover. LINA [12] is another ID/locator split-based mobility protocol. Here, IDs of 128-bit length are formed by concatenating location-independent prefixes (of 64 bits) and node IDs (of 64 bits), while locators are formed by concatenating location-dependent network prefixes and node IDs. The network layer of the host protocol stack is divided into two sublayers: the identification sublayer and the delivery sublayer. The former carries out the ID-to-locator mapping function and the latter forwards packets using destination locators present in the packet header. It uses mapping agents to resolve IDs into corresponding locators. It is a host-based mobility approach, i.e., there is no support for network-based mobility and smooth handover. LISP [7] uses prefix aggregateable endpoint IDs (EIDs), which are also used as locators in the edge network. In the transit network, routing locators (RLOC) are used as locators. EIDs to RLOCs mapping takes place in the Ingress and Egress Tunneling Routers (ITR/ETR) located in the border between the edges and the transit network. LISP s main focus is to reduce the global Border Gateway Protocol (BGP) routing table size by using fewer RLOCs in the transit network. Since it also uses EIDs as local locators in edge networks, it does not provide host mobility support. To provide host-mobility, there is a proposal for having the host possess a lightweight version of the ITR/ETR functions [8]. However, it may not be effective for reducing the BGP routing table size, if a distinct RLOC is assigned to each host. Obviously, LISP lacks smooth handover functions. Considering the limitations of related work, we recently proposed the HIMALIS architecture [3] for the new generation network. Figure 1 shows the architectural components and the protocol stack in the HIMALIS architecture. The end hosts as well the border routers or gateways have the identity layer inserted between the network and transport layers. The identity layer executes mobility functions when it receives mobility indications from the network layer. The mappings between hostnames, IDs and locators are stored in two different registries: Domain Name Registry (DNR) and Host Name Registry (HNR). That is, these registries are used to resolve hostnames to IDs and locators during a communication initialization phase. The border routers connecting edge networks to the global transit network also cache ID-to-locator mapping data in their ID tables. The border routers distribute ID-to-locator-mapping updates in the event of host mobility. This architecture supports a hybrid of host- and network-based mobility. The end nodes as well as network nodes, such as border routers, maintain host ID-to-locator mapping caches, participate in mobility signaling and use host IDs present in the identity header as the reference value to dynamically change locators present in the network layer header, while keeping the change hidden from the transport and application layers. Thus, the architecture is mobility friendly. In 5 Domain Name Registry (DNR)/ Host Name Registry (HNR) Global Transit Network Core Routers Border Router (BR) Border Router (BR) Hosts Host Edge Network Edge Network Host (a) Network components layout Host Application Application Transport Border Router Border Router Transport Identity Network Identity Net. Net. Core Router Network Identity Net. Net. Identity Network Edge Network Global Transit Network (b) Protocol stack layout Edge Network Figure 1: ID/locator split architecture components and protocol stack fact, HIMALIS provides other support, such as for heterogeneous network layer protocols and resource constrained nodes. We refer the interested reader to [3] for additional descriptions of the architecture. 3 Mobility Related Functions in ID/Locator Split Architectures In this section, we discuss the different types of mobility-related functions available in the ID/locator split architectures. 3.1 Name Resolution System Mobility support requires the name resolution system to have functions not only for faster resolution of hostnames to locators but also for faster update of the records in the event of changing locators when 6 mobile hosts move. The Internet employs the DNS to resolve domain names into IP addresses and other resources. Internet applications resolve the domain name into an IP address during a communication initialization phase via a DNS record lookup process. Although DNS has been providing scalable and faster name resolution, it is not suitable for fast updating of the hosts dynamic information, because of the existence of multiple cached copies in the global DNS server system. For efficient mobility support, in addition to DNS, a new mapping system is needed to store hosts dynamic information, such as locators. In fact, the new mapping system would work as the fixed anchor point, where mobile hosts reachability information is stored and updated. The rendezvous servers in HIP and host name registries in HIMALIS serve as the fixed anchor points. In HIP [11], mobile hosts domain names and their static rendezvous server s locators are registered with the DNS, while the mobile host s locator is stored in the rendezvous server. When a correspondent host wants to communicate with the mobile host, the correspondent host resolves the mobile host s fully qualified domain name (FQDN) into the host ID and locator by sending a name resolution query to a DNS server. The host locator received in the response is in fact the locator of the rendezvous server. Therefore, the first packet sent by the correspondent host to the mobile host s ID and locator goes to the rendezvous server, which searches its database for the mobile host s locator to relay the packet to the mobile host s current location. When the mobile host changes its locator due to mobility, it sends a location update request to the rendezvous server. Thus, the rendezvous servers are somewhat similar to the home agents in Mobile IP. Similarly, in LISP [7], mapping servers [9] are used to store dynamic mapping between host IDs and locators. LISP mapping servers provide ID-to-locator mapping records to ITRs. Unlike HIP s rendezvous servers, LISP mapping servers do not receive and relay the first data packet. HNR Records mh#domain-1.com MH s ID & Loc HNR DNR Records domain-1.com HNR s ID & Loc Mobile Host (MH) (mh#domain-1.com) 6 4 Query Relayed 5 Data communication Hostname Lookup MH s ID &Loc 3 Query: mh#domain-1.com? 1 Query: domain-1.com? DomainNameLookup 2 Response: HNR s ID & Loc DNR Correspondent Host (CH) (ch#domain-2.com) Figure 2: Hostname resolution process In HIMALIS [3], the HNRs are the fixed anchor points that store the locators of mobile hosts, whereas the HNRs locators are stored in the DNR. To make the hostname resolution process efficient, the for- 7 mat of the global hostname is changed slightly. In the current Internet, the domain name and hostname semantics are mixed together. In contrast, the HIMALIS architecture separates the global domain name and local hostname by using a hash # sign. That is, a global hostname is formed by concatenating its local hostname and global domain name with the # sign. For example, if a mobile host with a local hostname kafle-pc is administratively associated with a domain nict.go.jp, its global hostname would be kafle-pc#nict.go.jp. In the communication initialization phase, a correspondent host resolves the mobile hosts hostname into the host ID and locator in two steps, as shown in Figure 2. Suppose the correspondent host (hostname: ch#domain-2.com) wants to communicate with the mobile host (hostname: mh#domain-1.com). The correspondent host first sends a domain name resolution request to a DNR and gets the HNR s ID and locator. It then sends a hostname resolution request to the HNR. The HNR searches its database for the record containing the mobile host s ID and locator. The HNR may have two choices in sending the host ID and locator mapping record: (1) it may send the record directly to the correspondent host, or (2) it may relay the hostname resolution request to the mobile host. Figure 2 depicts the second choice, which may have the following advantages: 1. If the mobile host is multihomed to different networks and has several locators, it can select the most appropriate locator for the communication with the correspondent host. The network resource availability and intended service requirements (if the name resolution request also includes information about the type of communication service the correspondent host wants to have with the mobile host) can be used as decision parameters in the locator selection process. 2. HNR record update frequency of a multihomed host can be reduced by registering only a locator belonging to the network that has the widest coverage (e.g., a locator associated with a cellular network or an explicit signaling network such as that used in the MIRAI architecture [19]). The locators associated with other networks are not registered in the HNR. For example, a mobile host with both cellular and wireless LAN interfaces would require only registration of the locator associated with the cellular network. 3.2 Mobility Functions in Hosts and Network Nodes Host-based mobility support requires the host protocol stack to possess functions to carry out the following tasks: detect movement, configure a new locator, carry out location update signaling, use the new locator in data packet headers, and hide the locator change from the upper layers. The mobility detection and new locator configuration tasks are usually performed by the network layer protocols. The remaining tasks are performed by the mobility specific functions located in Layer 4 (i.e., in the identity layer in HIP and HIMALIS) or in Layer 3.5 (in LINA and Mobile IP). Since the inclusion of ID/locator split functions in the host protocol stack naturally enables hosts to detect movement and carry out mobility-related signaling, pure network-based mobility support (such as in Proxy Mobile IP, where hosts do not carry out any mobility related functions), is not required in ID/locator split architectures. Similarly, ID/locator mapping functions installed in border routers or gateways for making the core routing scalable or providing traffic engineering (as proposed in LISP and HIMALIS) are also helpful to handover optimization, mobility across heterogeneous network protocols, and route optimization. 3.3 Handover Optimization Handover optimization functions aim at reducing one or more of the following three parameters: handover delay, packet loss during handover, and signaling traffic due to handover. Among the ID/locator split-based architectures, only HIMALIS possesses handover optimization functions. These functions 8 are distributed both in the end hosts and in the border routers, which exchange the mobile host s locator update signalin
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks