Elsevier

Computer Communications

Volume 31, Issue 11, 15 July 2008, Pages 2739-2754
Computer Communications

Organizational virtual zones: Control of Internet edges using multi-level flat identifiers

https://doi.org/10.1016/j.comcom.2008.03.028Get rights and content

Abstract

The current Internet architecture neither supports end-device mobility nor network mobility. Moreover, it does not properly handle the multi-homing as each interface of a multi-radio equipped terminal generally appears as a completely different node to the core network. Initiatives for splitting the locator and identity of a node are currently under analysis but there is still much work to be done concerning the architectural aspects in order to provide a scalable routing mechanism based on node identifiers that belong to a flat address space. This paper presents a new architecture encompassing both a global look-up system based on multi-level flat identifiers and a modular routing architecture on the edges based on inter-network path-capability information exchange. This routing architecture allows the aggregation of traffic of the same class and provides QoS guarantees across heterogeneous wireless–wired edge networks. A similar modular approach proposed for the Internet core inter-works with the border gateway protocol without changing it. The proposed mechanism exploits the locator-identity split and the co-location of heterogeneous network domains that belong to the same organization (access provider to the Internet) by creating a virtual organizational zone. The virtual zone allows a transparent control of mobility and multi-homing, hiding them from the Internet core and from the user. It offers access providers (mobile or not) the possibility of implementing load balancing, multi-radio switching and multi-radio transmission diversity, while keeping the current Internet core intact.

Introduction

While the discussions on the necessary changes to the current Internet architecture address a wide range of issues, including QoS guarantee, mobility, routing and security, there is at least a consensus on the need of splitting the locator and the identity of a node for the future Internet architecture. The current IP architecture was originally designed for fixed hosts and networks and is not well suited to deal with the increasing presence of moving networks and mobile devices. The advantages brought by a clear locator-identity splitting become truly clear upon considering the quick expansion and the future trend of wireless communication systems. The hierarchical structure of the current IP architecture allows B order Gateway Protocol (BGP) routers [7] to aggregate information about distant networks and thus to reduce the number of routing table entries. Although it is this hierarchical IP numbering that makes the Internet scalable, it is precisely this hierarchical addressing structure – implicitly attached to a location – which makes mobility of networks and devices so difficult to handle. The problem stems from the fact that in the current Internet architecture any IP address is always associated with a determined location. Thus, when the location of a host changes its IP address must change accordingly. The IP renumbering problem related to mobile communications will certainly get worse in the future as along with networks operating in vehicles such as cars, trains and airplanes, there is also a predicted increase of sensor networks, robots and other types of devices that may vastly increase the amount of devices connected to the Internet. Projects carried out in IETF to provide mobility-support extensions to the current IP architecture such as Mobile IP [7], MOBIKE [8], NetLMM [9] and NEMO [10] only provide add-on solutions without addressing the main structural problem of the architecture, which is the simultaneous use of an IP address as both the locator and the identifier of a node. The most recognized and widely known proposal that supports the locator-identifier split architectural concept is the host identity protocol (HIP) [11], which has set ground for further work. In our opinion, one of the limitations of HIP is that the locator-identifier split is done only on the end-nodes. However, the future Internet architecture should consider the split of identifiers and locators on any discontinuity network point, such as an interface between two heterogeneous networks [12]. Thus, every single network with a consistent internal routing mechanism (associated with the term locator domain defined in Section 2) is provided with location privacy and a clear location-identity splitting.

The dissociation of locators and identifiers is also important to allow the scalability of IPv4 and IPv6. For routing to scale, locators need to be assigned according to a topology and most likely change as topology changes. Nowadays, assignment of IP addresses to communication sessions restricts the use of the multi-homing capability of end-devices, since the session is constraint to the characteristics – which can evolve over time – of the interface configured with such IP address. The same applies to networks, since the assignment of IP addresses to organizations is hard due to the same topological constraints on the inter-network links in particular in the case of moving networks. Hence organizations should also have self generated ‘identifiers’ so that they do not have to renumber their interfaces if they change providers or become multi-homed within the network topology.

Node identifiers should have a persistent nature in order to create a long-term binding to the node that they are naming. The identifiers should not change during the long-lived network sessions. To use long-term identifiers in dynamic network topologies, identifiers should be globally unique and be self generated. Hence, identifiers must remain the same when an organization or single device changes provider or otherwise moves to a different point in the network topology. Finally, identifiers must hide the diversity of interfaces that may be used by some nodes. This is, the communication between nodes must be set up and maintained without the nodes having to know which of the peer’s interface is more suitable for the new or on-going communication sessions. That information should be purely handled on the network side and not on the terminal side. This is achieved by our proposed architecture through the grouping of heterogeneous networks that are co-located in a certain geographical region. This feature allows the system to hide mobility events and multi-homing from the Internet core and the communication partner zone. In other words, mobility events and multi-homing aspects are kept inside the organizational zones’ administration and are not reflected out of the respective zone.

The rapid increase of 3G mobile system’s users as well as the quick expansion of WLAN hot spots and adoption of new systems like WiMax around the world, gives us a hint that future wireless communication systems will be based on a co-existence of a number of wireless radio technologies and mobile operators. In the future, mobile operators will most likely offer their users a great variety of services such as voice, video and data applications running on multi-radio terminals. Ideally these mobile terminals will be capable of simultaneously using multiple radio interfaces from one or more mobile access providers and keep the technical communication aspects hidden from users as much as possible. In the current Internet architecture generally each of these radio interfaces has a different IP address and, as a result, is treated independently even if managed by the same mobile access provider. Hence from the core network’s perspective a single multi-radio node may appear as a set of completely independent nodes with non-correlated traffic In contrast our proposal allows every node to have a single identity from the core network perspective. To this end we apply two basic network design criteria: (i) the use of a unique identifier for a given mobile node, irrespective of its number of wireless interfaces or access providers, and (ii) the grouping of access networks that are geographically co-located irrespective of their technical nature or owner. Such re-structuring at the edges of the Internet may simplify the creation in the future of new functionalities such as multi-radio switching and transmission diversity as well as load balancing.

From a broader network perspective, in our opinion multi-radio moving nodes will be part of a future communication environment that will encompass the current Internet as a core packet-switching network and an increasing number of heterogeneous and dynamic access networks around it. These edge networks are self-contained, have coherent internal addressing and routing schemes, and dynamic external connectivity with their neighbours and with the Internet core. Some of these edge networks can be mobile, although their mobility pattern most likely will not be as dynamic and random as the one evidenced by moving nodes, which will exist in a huge number.

In this paper, we present a new architecture to re-structure the edges of the Internet based on organizational virtual zones. The architecture encompasses a new modular routing system in which a dynamic control of inter-network quality-of-service negotiation mechanism, called inter-network QoS agreements (INQA) [3], is placed on top of different routing protocols operating in heterogeneous access networks. The novel reorganization of the edges of the Internet as well as the path-capability aware routing mechanism helps organizations to provide end-to-end QoS across heterogeneous networks as well as to deal more efficiently with multi-homing and mobility. The fundamental concept can be applied to any existent architecture, including the Node ID architecture [12] and the current Internet architecture. Although the proposed architecture utilizes INQA and the Node ID architecture for path-capability-aware routing and for the identity-locator split in every end-node and border router of the network, both must be regarded as examples for components instead of an integral part of the architecture.

The paper is structured as follows. Section 2 gives an overview of the Node ID architecture and the INQA protocol. Section 3 describes the proposed organizational virtual zones’ architecture. Section 4 describes the look-up phase and the communication set-up. The capability-aware routing mechanism is introduced in Section 5. Section 6 presents some related work. Finally, some concluding remarks are made in Section 7.

Section snippets

Node ID architecture

The Node ID architecture proposed in the framework of the Ambient Networks project [12] provides a locator-identity splitting by allowing each node in the Internet, and not only an end-node, to self generate a cryptographic identifier, which can be the result from hashing a public key or other types of cryptographic material. The identifier label of a node is sequentially registered up to the core network in a path that crosses different locator domains (LDs) so as to make the node globally

Extension of the Node ID architecture using organizational virtual zones

One of the major concerns regarding the Node ID architecture is the scalability of the network as all nodes must register its identifier (ID) in the Internet core. As the number of mobile devices is expected to increase very rapidly during the coming years, the total number of nodes that may be simultaneously registered in the Internet core could reach the order of billions in a global scale. In [13] the authors try to tackle this scalability problem by assigning a cryptographic identifier to

Look-up phase and communication set up

A communication is initiated by the look-up phase when a given node intends to connect to any other node in the Internet. Before establishing the communication the source node has to obtain both the NID and Zone ID of the destination node using only its FQDN. Let us suppose that the FQDN of the destination node is Paul_PDA.docomo.com. and the source node sends a query including the FQDN to a hierarchical name resolution system like domain name server (DNS), which replies by sending back to the

Routing mechanism

In the organizational-zone architecture, communications are done based on a three tuple 〈NID_dst, ZoneID_dst, SLS〉 based on which a source NID can send packets to a destination NID. Packets sent by the source may traverse three distinct routing areas: (a) intra-LD routing in the source and destination LDs; (b) inter-LD routing in the source and destination zones; (c) BGP in the IP core, as illustrated in Fig. 8 (following the example of Fig. 4).

The proposed routing system is built on top of the

Related work

The first question that we can pose, is how to handle the mobility of a very large number of moving nodes (in the order of billions), being those nodes known by their flat address-space identifier. Any solution that encompasses the registration of the nodes location in the Internet core may suffer from scalability problems. In [13] the author tries to tackle this scalability problem by assigning a cryptographic identifier to every locator domain (LD), namely LD_ID, and then registering these

Conclusions and future work

Capability-aware routing deals with routing mechanisms that can establish and support end-to-end communications across a dynamically composing topology of wired and wireless networks. In this paper a new Internet architecture based on organizational virtual zones that exploits the regional co-location of different locator domains has been presented. The scheme allows organizations or operators hiding mobility from the core network’s perspective, handling multi-homing in a more efficient way and

References (19)

  • J. Babiarz, K. Chan, F. Baker, Configuration guidelines for diffserv service classes, IETF, RFC 4594, August...
  • A. Doria, E. Davies, F. Kastenholz, Requirements for inter-domain routing, Internet Draft, October 2006,...
  • P. Mendes, J. Andres-Colas, C. Pinho, Information model for the specification of QoS agreements among ambient networks,...
  • N. Niebert et al.

    Ambient networks: an architecture for communication networks beyond 3G

    IEEE Wireless Communications

    (2004)
  • I. Psaras, L. Mamatas, P. Mendes, INQA: internetwork qos agreements – a new protocol for dynamic SLS control in next...
  • H. Schulzrinne, R. Hancock, GIST: General Internet Signaling Transport, IETF, Internet Draft, August 2006,...
  • C. Perkins, IP Mobility Support for IPv4, IETF RFP 3344,...
  • P. Eronen, IKEv2 Mobility and Multihoming Protocol, IETF draftietf-mobike-protocol-08.txt,...
  • G. Giaretta, et al., NetLMM Protocol, IETF draft-giaretta-netlmmdt-protocol-00.txt,...
There are more references available in the full text version of this article.

Cited by (0)

1

This work was done when the author was with D.C.M. Euro-Labs.

View full text