1 Introduction
2 Cellular Internet
3 Protocol Overview
3.1 Features
3.2 Routing
3.3 Handoff
3.4 Paging
References
1 Introduction
The development of commodity-based palmtop devices with
built in high-speed packet radio access to the Internet will have a major impact
on the mobile telecommunications industry and the way we communicate. The
availability of cheap, ubiquitous and reliable wireless Internet access will
shift the service base traditionally found in mobile telecommunication networks
toward emerging wireless Internet Service Providers (ISPs). This will result in
significant demands being placed on both existing and next-generation cellular
and IP networks.
In this paper, we present an overview of Cellular IP [4],
an Internet host mobility protocol [5] that takes an
alternative approach to that found in mobile telecommunications (e.g., General
Packet Radio Service [1]) and in IP networking (Mobile IP [2]).Cellular
IP represents a new mobile host protocol
that is optimized to provide access to a Mobile IP enabled Internet in support
of fast moving wireless hosts. Cellular IP incorporates a number of important
cellular principles but remains firmly based on IP design
principles
allowing Cellular IP to scale from pico to metropolitan area installations.
The paper is structured as follows. In Section 2, we
discuss the concept of an IP-based Cellular Internet where Cellular IP provides
micro-mobility support interworking with Mobile-IP which provides macro-mobility
support between Cellular IP wireless access networks. Following this in Section
3 we present an overview of the Cellular IP routing, handoff and paging
algorithms. The Cellular IP distributed location management and routing
algorithms lend themselves to a simple, efficient and low cost implementation
for host mobility requiring no new packet formats, encapsulation or address
space allocation beyond what is present in IP. In Section 4
we provide some concluding remarks.
2 Cellular Internet
Recent initiatives to add mobility to the Internet
mostly focus on the issue of address translation [3]
through introduction of location directories and address translation
agents. In these protocols (e.g., Mobile IP) packets addressed to a mobile host
are delivered using regular IP routing to a temporary address assigned to the
mobile host at its actual point of attachment. This approach results in simple
and scalable schemes that offer
global mobility. Mobile IP
is not appropriate, however, for fast mobility and smooth handoff because after each
migration a local address must be obtained and communicated to a possibly
distant location directory or home agent (HA).
Cellular mobile telephony systems are based on a different
concept from that of Mobile IP. Instead of aiming at global mobility support,
cellular systems are optimized to provide fast and smooth handoff within
restricted geographical areas. In the area of coverage, mobile users have
wireless access to the mobility unaware global telephony network. A scalable
forwarding protocol interconnects distinct cellular networks to support roaming
between them.
Even in limited geographical areas, however, the
number of users can grow to a point where using fast lookups for per user data
bases is no longer viable. In addition, mobility management requires mobile
hosts to send registration information after migration. The resulting signaling
overhead has significant impact on the performance of wireless access networks.
To overcome this problem, cellular telephony systems require mobile hosts to
register after every migration only when they are engaged in `active' calls. In
contrast, 'idle' mobile hosts send registration messages less frequently and as
a result can roam large areas without loading the network and the mobility
management system. In this case, the location of idle mobile hosts is only
approximately known to the network. To establish a call to an idle mobile host,
the mobile host must be searched for in a limited set of cells. This feature of passive
connectivity allows cellular networks to accomodate a very large number of
users at any instance without overloading the network with large volumes of
mobility management signaling information.
Cellular networks offer a number of desirable features
which if applied correctly could enhance the performance of future wireless IP
networks without loosing any of the important flexibility, scalability and
robustness properties that characterize IP networks. However, there are
fundamental architectural differences between cellular and IP networks that make
the application of cellular techniques to IP very challenging. Cellular
telephony systems rely on the restrictive "circuit'' model that requires
connection establishment prior to communication. In contrast, IP networks
perform routing on a per packet basis. In addition, current cellular
systems are strictly based on hierarchical networks based on costly
mobile-aware nodes (e.g., MSC). We believe that a future ``Cellular Internet''
should be based on IP, inheriting its simplicity, flexibility and robustness. A
Cellular Internet should leverage mobility management and handoff techniques
found in cellular networks. A single scalable host mobility protocol should be
capable of flexibly supporting pico, campus and metropolitan area networks based
on a set of simple and cheap network nodes that can be easily interconnected to
form arbitrary topologies and operate without prior configuration.
3 Protocol Overview
In what follows, we provide an overview of the Cellular IP
features and algorithms; that is, the Cellular IP routing, handoff and paging
algorithms.
For a full discussion, specification and evaluation of Cellular IP see [4],
[5], [6], respectively.
3.1 Features
Cellular IP inherits cellular systems principles
for mobility management, passive connectivity and handoff control, but is
designed based on the
IP paradigm. The universal component of a Cellular IP network is the base
station which serves as a wireless access point but at the same time routes
IP packets and integrates cellular control functionality traditionally found in
Mobile Switching Centers (MSC) and Base Station Controllers (BSC). The base
stations are built on regular IP forwarding engines, but IP routing is replaced
by
Cellular IP routing and location management. The Cellular IP network is
connected to the
Internet via a gateway router. Mobility between gateways (i.e., Cellular
IP access networks) is managed by Mobile IP while mobility within access
networks is
handled by Cellular IP. Mobile hosts attached to the network use the IP address
of the gateway as their Mobile IP care-of address. Figure~1 illustrates the path
of the packets addressed to a mobile host. Assuming Mobile IPv4 [2]
and no route optimization [7], packets will be first
routed to the host's home agent and then tunneled to the gateway. The gateway
"detunnels'' packets and forwards them toward base stations. Inside the
Cellular IP network, mobile hosts are identified
by their home addresses and data packets are routed without tunneling or
address conversion. As discussed later, the Cellular IP routing protocol ensures
that packets are delivered to the host's actual location. Packets transmitted by
mobile hosts are first routed to the gateway and from there on to the Internet.
Figure 1: Cellular IP Access
Network
In Cellular IP, location management and handoff support are
integrated with routing. To minimize control messaging, regular data packets
transmitted by mobile hosts are used to
establish host location information. Uplink
packets are routed from mobile to the gateway on a hop-by-hop basis. The path
taken by these packets is
cached in base stations. To route downlink packets addressed to a mobile host the path
used by recent packets transmitted by the host is reversed. When the mobile host has no data to
transmit then it periodically sends empty IP packets to the gateway to maintain
its downlink routing state. Following the principle of passive connectivity
mobile hosts that have not received packets for a certain period of time allow
their downlink soft-state routes to timeout and be cleared from the routing
cache.
In order to route packets to idle hosts a Cellular IP
mechanism called paging is used.
3.2 Routing
The Cellular IP gateway periodically broadcasts a beacon
packet that is flooded in the access network. Base stations record the interface
they last received this beacon through and use it to route packets toward the
gateway. All packets transmitted by mobile hosts regardless
of their destination address are routed to the gateway using these routes.
As these packets pass each node on route to the
gateway their route information is recorded as follows. Each base station
maintains a routing cache.
When a data packet originated by a mobile host enters a base station the local
routing cache stores
the IP address of the source mobile host and the interface over which the packet
entered the node. In the scenario illustrated in Figure1 data packets are
transmitted by a mobile host with IP address X and enter BS2
through its interface a. In the routing cache of BS2
this is indicated by a mapping (X,a). This mapping remains valid for a
system specific time route-timeout and its validity is renewed by each
data packet that traverses the same interface coming from the same mobile. As
long as the mobile host is regularly sending data packets, base stations along
the path between the mobile host's actual location and the gateway maintain
valid entries in their routing cache forming a soft-state route between the
mobile host and gateway nodes. Packets addressed to the same mobile host are
routed on a hop-by-hop basis using the established routing cache.
A mobile host may sometimes wish to maintain its routing cache mappings even
though it is not regularly transmitting data packets. A typical example for this
is when the host is the receiver of a stream of UDP packets and has no data to
transmit. To keep its routing cache
mappings valid the mobile host transmits route-update packets at regular
intervals
called route-update time. These packets are empty data packets addressed
to the gateway. Route-update packets have the same
effect on routing cache as normal data packets; however, they do not leave
Cellular IP access networks.
3.3 Handoff
The Cellular IP hard handoff algorithm is based on a
simplistic approach to mobility management that supports fast and simple handoff
at the price of potentially some packet loss. Handoff is initiated by mobile
hosts. Hosts listen to beacons transmitted by base stations and initiate handoff
based on signal strength measurements. To perform a handoff a mobile host has to
tune its radio to the new base station and send a route-update packet. This
creates routing cache mappings on route to the gateway hence configuring the
downlink route to the new base station. Handoff latency is the time that elapses
between the handoff and
the arrival of the first packet through the new route. For hard handoff this
equals the round-trip time between the mobile host and the cross-over point
which is the gateway
in the worst case. During this time, downlink packets may be lost. The mappings
associated with the old base station are not cleared at handoff, rather, they
timeout as the associated soft-state timers expire.
Before the mappings timeout, a period exists when
both the old and new downlink routes are valid and packets are delivered through
both base stations. This feature is used in the Cellular IP semisoft handoff
algorithm that improves handoff performance but still suits the lightweight
nature of the base protocol providing probabilistic guarantees instead of fully
eliminating packet loss. Semisoft handoff adds one additional state variable to
the existing mobile state maintained
at mobile hosts and base stations. The semisoft handoff procedure has two
components. First, in order to reduce handoff latency, the routing cache
mappings associated with the new base station must be created before the actual
handoff takes place. When the mobile host initiates a handoff, it sends a semisoft
packet to the new base station and immediately returns to listening to the
old base station. While the host is still in contact with the old base station,
the semisoft packet configure routing cache mappings associated with the new
base station. After a semisoft
, the host can perform a regular handoff. The semisoft delay
can be an arbitrary value between
the mobile-gateway round-trip time and the route-timeout. The delay ensures
that by the time the host tunes its radio to the new base station, its downlink
packets are delivered through both the old and new base stations.
While the semisoft packet ensures that the mobile
host continues to receive packets immediately after handoff, it does not,
however, fully assure smooth handoff. Depending on the network topology and
traffic conditions, the time to transmit packets from the cross-over point to
the old and new base stations may be different and the packet streams
transmitted
through the two base stations will typically not be synchronized at the mobile
host. If the new base station "lags behind'' the old base station, the
mobile host may receive duplicate packets. Reception of duplicate packets in
this case is not disruptive to
application operations. If, however, the new base station "gets ahead''
then packets will be deemed to be missing from the
data stream observed at the receiving mobile host. The second component of the
semisoft handoff procedure is based on the observation that perfect
synchronization of the two streams is not necessary. The condition can be
eliminated by temporarily introducing into the new path a constant delay
sufficient to compensate, with high probability, the time difference between the
two streams. This can be best achieved at the cross-over switch that understands
that a semisoft handoff is in progress due to the fact that a semisoft packet
has arrived from a mobile host that has a mapping to another interface.The mapping created by the
semisoft packet has a flag to indicate
that downlink packets routed by this mapping must pass a "delay device''
before transmission. After handoff, the mobile host will send data or
route-update packets along the new path which will clear this flag and cause all
packets in the delay device to be forwarded to the mobile host.
3.4 Paging
Cellular IP defines an idle mobile host as one that
has not received data packets for
a system specific time active-state-timeout. In this respect, idle
mobile hosts allow their respective
soft-state routing cache mappings to time out. These hosts transmit paging-update
packets
at regular intervals defined by paging-update-time. The paging-update
packet is an empty IP packet addressed to the gateway that is distinguished from
a route-update packet by its IP type parameter. Paging-update packets are sent
to the base station
that offers the best signal quality. Similar to data and route-update packets,
paging-update packets are routed on a hop-by-hop basis to the gateway. Base
stations may optionally maintain paging cache. A paging cache has the
same format and operation as a routing cache except for two differences. First,
paging cache mappings have a longer timeout period called paging-timeout.
Second, paging cache mappings are updated by any packet sent by mobile hosts
including paging-update packets. In contrast, routing cache mappings are updated
by data and route-update packets sent by mobile hosts. This results in idle
mobile hosts having mappings in paging caches but not in routing caches. In
addition, active mobile hosts will have mappings in both types of cache. Packets
addressed to a mobile host are normally routed by routing cache mappings. Paging
occurs when a packet is addressed to an idle mobile host and the gateway or base
stations find no valid routing cache mapping for the destination. If the base
station has no paging cache, it will forward the packet to all its interfaces
except for the one the packet came through. Paging cache is used to avoid
broadcast search procedures found in cellular systems. Base stations that have
paging cache will only forward the paging packet if the destination has a valid
paging cache mapping and only to the mapped interface(s). Without any paging
cache the first packet addressed to an idle mobile host is broadcast in the
access network. While the packet does not experience extra delay it does,
however, load the access network. Using paging caches, the network operator can
restrict the paging load in exchange for memory and processing cost. Idle mobile
hosts that receive a packet move from idle to active state, start their
active-state-timer and immediately transmit a route-update packet. This ensures
that routing cache mappings are established quickly potentially limiting any
further flooding of messages to the mobile host.
References
[1] G. Brasche,
B.Walke, ''Concepts, Services and Protocols of the New
GSM Phase 2+ General Packet Radio Service,'' IEEE Communications
Magazine, pp.94-104, August 1997.
[2] Charles Perkins,
editor, ``IP Mobility Support,'' Internet RFC 2002,
October 1996.
[3] Pravin Bhagwat,
Charles Perkins, Satish Tripathi, ``Network Layer Mobility: an
Architecture and Survey,'' IEEE Personal Communications Magazine, Vol.3,
No.3,
pp.54-64, June 1996.
[4] Andras G. Valko,
"Cellular IP: A New Approach to
Internet Host Mobility,''
ACM Computer
Communication Review, January 1999.
[5] A.Valko, A.
Campbell, J. Gomez, "Cellular
IP'', Internet Draft,
draft-valko-cellularip-00.txt, November 1998,
Work in Progress.
[6] A.Valko, J.
Gomez, S. Kim, A. Campbell, "Performance
of Cellular IP Access Networks,'' Proc.
of 6th IFIP International Workshop on Protocols for High Speed Networks
(PfHSN'99), Salem, August 1999.
[7] David B.
Johnson, Charles Perkins, "Route Optimization in Mobile
IP,'' Internet Draft,
draft-ietf-mobileip-optim-07.txt, November 1998, Work in Progress.
[8] D. Clarc, J.
Wroclawski, "An Approach to Service Allocation in the
Internet,'' Internet Draft,
draft-clarc-diff-svc-alloc.00.txt, July 1997, Work in Progress.
[9] The
Cellular IP Project at Columbia University http://comet.columbia.edu/cellularip
