Document Sample
					ABSTRACT:                                        during the execution of its data service. To

     This paper presents the fault tolerance     avoid     interrupting    the   ongoing     data

of Mobile IP in wireless systems. Mobile         session, The Internet Engineering Task

IP can support wireless users with               Force (IETF) has defined Mobile IP as one

continuous network connections while             which enables wireless users to maintain

changing locations. It is achieved by            ongoing data sessions without interruption

allocating a number of mobility agents           while changing locations.

(foreign agents and home agents) in the          To provide the functionality of Mobile IP

architecture of a wireless system. If a          in wireless system a number of mobility

failure occurs in a mobility agent, the          agents are required in the architecture of

wireless users located in the coverage area      the wireless system. The mobility agents

of the faulty mobility agent will lose their     are classified into two types: Home agent

network connections. To tolerate the             (HA), foreign agent (FA).The FA logically

failures of mobility agents, this paper          connects     with   several     radio     access

presents    an      efficient   approach    to   networks (RANs) to form a wireless data

maintaining the network connections of           serving area. The HA maintains the

wireless users without being affected by         current       addresses         of      mobile

the failures. Once detecting a failure in a      nodes(MNs).When an MN stats a data

mobility agent, failure-free mobility agents     session, a data request is first sent to the

are dynamically selected to be organized         located RAN of the MN, then to a serving

as a backup set to take over the faulty          FA and, finally ,to a corresponding

mobilility agent.                                application server. The application server
                                                 processes the data requests and sends back

1. INTRODUCTION                                  the packets of the handling results to the
                                                 MN. The packets send to the MN will be
       Due to the rapid progress of
                                                 first     intercepted     by     the      MN’s
wireless communication technology, there
                                                 corresponding HA. The HA looks up the
is a growing demand for accessing data by
                                                 current address of the MN and tunnels the
wireless    systems      .Mobility   is    one
                                                 packets to the FA serving the MN. Then,
important    characteristic     of   wireless
                                                 the FA detunnels the packets and forwards
systems .Each wireless user (mobile node)
                                                 then to the MN. From this scenario of a
may change its location several times
wireless data session, we can know that, if     be the backup members of the faulty HA.
a failure in a FA, all MNs located in its       The selected failure-free HAs, instead of
data serving area cannot perform wireless       the faulty HA, intercept the packets
data sessions again. Likewise, if an HA         moving toward the faulty HA, and then
crashes, the response packets of the data       send the packets to corresponding MNs.
session can not be sent back to the
corresponding MN.                               2. BACKGROUND:
This paper presents an efficient approach              This     section     describes   the
to providing fault-tolerant capability in the   background materials of this paper. Firstly,
wireless    system     with    Mobile     IP    an overview of the Mobile IP is given.
functionality.                                  Then the system model is described. Next,
     The proposed approach is based on          the fault assumption is made. Finally,
the concept of resource sharing to redirect     related work is reviewed.
the work loads of faulty FA(HA)to other
failure-free FAs(HAs).Unlike the previous       2.1 MOBILE IP Overview:
approach, the fault tolerant method of an            Mobile IP is designed to support node
FA in the proposed approach is different        mobility by using two types of mobility
from that of an HA. Once detecting a            agents: home agent (HA) and foreign
failure in a FA, one or more failure-free       agent (FA).Initially, each mobile node
FAs are dynamically selected to be              (MN) has a unique address to be managed
organized as a backup set for the faulty        by an HA on its home network. If an MN
FA. Then, a system-initiated handoff is         moves from its home network to foreign
issued to virtually move the MNs now            network, a “care-of-address” (COA) is
served by the faulty FA to the serving          allocated to the MN to reflect the MN’s
areas of the FAs in the backup set. The         current point of attachment. The FA on the
data executable capabilities for the MNs        foreign network also adds an entry
can be continuously supported by the            corresponding to the MN in the visitor list.
failure-free FAs in the backup set. For the     Then, a register message is sent to the
fault tolerance of an HA, if a failure is       MN’s HA to create or modify the MN’s
detected in an HA, one or more failure-         mobility binding which associates the
free HAs are also dynamically selected to       MN’s home address, the located FA, and
the COA. Packets from a correspondent           to forward packets .In addition ,between
host (CH) to an MN or addressed to the          RAN’s         and    FA’s    there   is   an
home address of the MN. If the MN is            interconnection network to deliver the
away from its home network, the HA on           wireless data requests from MN’s to the
the home network can intercept the              core network and to send back the
packets and tunnel them to the located FA       response packet from the core network to
of the MN. Then, the FA decapsulates the        MN’s. The interconnection network can be
packets and forwards them to the MN.            a frame relay network, ATM network, or
                                                IP network.
2.2 SYSTEM MODEL:                               To manage all equipment in the wireless
            The system model considered in      system, operations, administration, and
this paper refers to the architecture of a      maintenance         (OA&M)    functions   are
third generation (3G) wireless system. The      necessary. The OA&M functions are
system      model   contains    three   major   classified into configuration management,
components:                                     fault         management,       performance
        Mobile node(MN)                        management, and security management.
        Radio access network(RAN)              The configuration management configures
        Core network                           equipment (e.g., RAN, FA, HA, AAA
MN interacts with RAN to obtain radio           server, router, etc.) with suitable resource
resources for performing wireless data          parameters. The performance management
sessions.                                       measures the resource utilization, loading
The RAN provides the transmission across        status, and other concerned values in the
the air interface. With the core network, it    equipment. The fault management is
is an IP based network and contains the         capable of detecting and reporting failures
following equipment:                            in      the    equipment.    The     security
        Foreign agent(FA)                      management monitors the access rights to
        Home agent(HA)                         the equipment.
        Intermediate routers
The FA and HA provide the wireless data         2.3 FAULT ASSUMPTION:
session with Mobile IP functionality. The                 In this paper, failures are only
intermediate routers assist FA’s and HA’s       assumed to occur in mobility agents. To
announce the presence of a mobility agent         3.1    FAULT        TOLERANCE              OF
(FA    or      HA),   the     mobility    agent   FOREIGN AGENT
periodically      transmits       an      agent             Upon detecting a failure in an
advertisement message on its located              FA, the MNs located in its serving area
subnet. When a failure occurs in a                will not be able to execute wireless data
mobility agent, the failure can be detected       sessions. The new arriving MNs are also
by not receiving an agent advertisement           affected. These MNs are called as
message within a period of time. The fail-        FA_failure-affected MNs. To resume the
stop assumption is also made so that the          data executive abilities of the FA_failure-
faulty mobility agent cannot send any             affected MNs, a system initiated handoff is
advertisement message again.                      performed to dynamically select multiple
 As an FA fails, its in-processing data           failure-free FAs to organize a backup set
requests will be lost. For a faulty HA, its       for the faulty FA. Then, the FA_failure-
in-processing response packets are also           affected MNs are virtually moved to the
lost. However, these lost data requests and       serving areas of the selected failure-free
response packets can be retransmitted by          FAs. Each of the selected failure-free FAs
an end-to-end reliable Transport Control          adds a number of visitor entries for the
Protocol (TCP).
3.            THE             PROPOSED
        This     section presents        a new
approach to tolerating the failures of
mobility agents. To avoid significant
performance degradation, the approach
dynamically selects multiple failure-free
mobility agents to form a backup set for
the faulty mobility agent. The work loads
of the faulty mobility agent are redirected
to the failure-free mobility agents in the
backup set.
                                            OA&M Center


                                            OA&M Network

                           R                                                HA
                           N                       FA
                                        2                           3   3
                           R                                       Core
                           A                                       Network
                           N                       FA
                                                                   3        3

                           N                       FA

                      1. Mobilenode                         RAN: Radio access network
                      2. Interconnection Network            FA: Foreign agent
                      3.Router                              HA: Home agent
                                                            OA&M: Operations Administration
                                                              and Maintenance
                            Fig. 1. Wireless data network model
                                                     FA_failure-affected         MNs    can   be
FA_failure-affected    MNs       that   have         continuously supported by their new
moved to it and informs those MN’s                   serving FAs. The workloads of the faulty
corresponding HAs of the new serving                 FA are redirected to other failure-free
FAs and the care-of-address to update                FAs.
their   mobility      bindings.         After         Originally, a handoff functions to
performing the system-initiated handoff,             switch an MN’s wireless communication
the network connections of the                       and maintains its network connection
                                                     when the MN moves from one radio
coverage   area   to   another.   In   the   backup     members.     Therefore,    the
proposed approach, the system-initiated      FA_failure-affected MNs are now served
handoff is to redirect the workloads of      by the backup members, but their
the faulty FA, but it does not introduce     locations are not changed (they still
location changes of the FA_failure           located in their respective radio coverage
affected MNs. This objective is attained     areas), as shown in the figure.2.
by modifying the relationship between
RANs and FAs, described as follows:
        As shown in the figure.1 there is
an interconnection network to connect
RANs with FAs. From the logic point of
view, there exists at least one delivery
path from an RAN to each FA.
Therefore, if an RAN receives a data
request from a MN, it can deliver the
data request to any FA to process it. The
RAN determines which FA to process its
received data requests based on its
internal FA-serving record. Initially, the
FA-serving record of an RAN is set to be
the identifier of a fixed FA. After
detecting a failure in an FA, one or more
failure-free FAs are selected to be the
backup members of the faulty FA. Then,
the FA-serving records of the failure
affected RANs (the RANs which have
an initial serving relationship with the
faulty FA) are reset as the identifiers of
the backup members. By modifying the
FA serving records, the serving FAs of
the failure-affected RANs become the
Fig. 2. Remapping the relationship
between RANs and FAs for fault



1. Mobile node      2, 3, 4. RAN       5. Faulty FA
6, 7. Failure-free FA             8. Interconnection

3.2       FAULT       TOLERANCE                 OF
                                                           the backup members of the faulty
          An HA assists a correspondent
                                                           HA.Then, the above three functions of
host (CH) to send packets to an MN
                                                           the faulty HA are restored on the backup
based on the following three functions:
                                                           members, described as follows:
          Mobility binding maintenance
          Packet interception
          Packet tunneling
If an HA fails, the HA_failure-affected
MNs (the MNs managed by the faulty
HA) will not be able to receive packets
from       CHs.To      resume        the    packet
receiving abilities of the HA_failure-
affected MNs, one or more failure-free
HAs are also dynamically selected to be
Fig .3. Packets destined to an HA-
failure-affected MN.

                                                        2          router

                                                  Wireless IP
                          R                       Backbone
             1                            F
                          A               A


                       : Original delivery path          : Fault-tolerant path
                        Of packets to a MN                    of packets to a MN

As mentioned in the section 2.1, when              Known. This points out that the mobility
an MN enters into a data serving area,             bindings of the faulty HA can be
the corresponding FA adds a visitor                restored by searching all the FA’s visitor
entry in its visitor list to record the            lists. The restored mobility bindings are
following information : the MNs data               then distributed to the backup members.
link-layer address, IP address, and home           With regard to the packet interception of
agent address. When the MN leaves the              the faulty HA, it restored on the backup
data serving area, its corresponding               members using the tunneling technique.
visitor entry in an FA’s visitor entry is          Usually, one or more routers are
made invalid. Therefore, if an MN has a            collocated with an HA on the same
valid entry in an FA’s visitor list, the           network segment to connect the HA with
MN is now located in the serving area of           the core network and the Internet. When
the FA. From the FA’s visitor lists, the           a CH sends packets to an MN, the packet
up-to-date locations of all MNs can be             is first received by the collocated
routers, and is then intercepted by the        Based on this loading information,
MN’s HA. After detecting a failure in an       multiple failure-free FAs (HAs) with
HA, its collected routers do not forward       low traffic are selected to be the backup
their received packets to the faulty HA.       members of the faulty FA (HA). Then,
Instead, the collected routers tunnel the      the   configuration           management      is
packets to the backup members, as              informed    to        configure   the    backup
shown in the fig.3.In such a case, a           members of the faulty FA (HA) by
packet from a CH to an HA-failure-             resetting appropriate parameters to some
affected MN is sent by twice tunneling.        equipment        in     the    core     network.
One is to let one backup member                Thereafter, the incoming workloads of
intercept the packet. Another is to tunnel     the faulty FA (HA) are redirected to its
the packet to the located FA of the            backup members.
HA_failure-affected MN. As for the
restoration of the packet tunneling, the
failure-free HAs selected as the backup
members, already possess the tunneling
        The     implementation    of    the
proposed approach can be integrated into
the OA&M.
As described in section 2.2, there are
four main functions in the OA&M.
When a failure in an FA (HA) is
detected, the failure event is sent to the
fault       management.     The        fault
management initiates the proposed fault-
tolerant approach for the FA (HA). At
first, it interacts with the performance
management to acquire the loading
status of the failure-free FAs (HAs).
4.1     IMPLEMENTATION                 OF
   To understand the implementation in
more detail, the fault-tolerant procedure
for the FA is given in Fig. 4.

Do not receive an agent advertisement message sent from a FA within a period of time
Send a FA failure event to the fault management
/*collecting the load status of each failure-free FA*/
Ask the performance management to respond with the loading status of each failure-free
/*calculating the number of FA-failure-affected MNs*/
FA-failure-affected_MNs 0 --number of FA_failure-affected MNs
For each RAN RANi managed by the faulty FA (each failure-affected RAN)
      n Number of associations for data sessions in the RAN
      FA_failure-affected_MNs FA_failure-affected_MNs+n
Fig. 4. The fault-tolerant procedure for
the FA. (a)Collecting the loading status               A MN issue a wireless data
of failure-free FAs and the number of          session,   a   corresponding     entry    is
FA_failure-affected MNs.                       generated in the RAN to record the
       There are three main tasks in the       association between the RAN and the
procedure. The first task is to collect the    FA serving the MN. The second task
loading status of each failure-free FA         selects multiple failure-free FAs to be
and calculate the number of FA_failure-        the backup members of the faulty FA, as
affected MNs, as shown in Fig 4a. The          shown in Fig.4b.
loading status of each failure-free FA         First, the failure-free FAs with lower
can be collected from the performance          traffic are selected. If the total resources
management of OA&M. For the number             selected failure-free FAs are insufficient
of FA_failure-affected MNs, it can be          to serve all failure-affected MNs, some
calculated by counting the number of           failure-free FAs with higher traffic will
association entries in the failure-affected    be selected to be the backup members. In
RANs. In an RAN’s coverage area, if an         addition, each selected failure-free FA
                                               also estimates the number of FA_failure-
affected MNs to be virtually moved to
its serving area based on the number of
available resources in it.

 MNs_assigned 0 --number of MNs already assigned their new serving FAs
 Lower_mark true
 For each failure-free FA2
     Available_resources[i] the number of available resources in the FA2
     Resources_to_use[i] 0 –number of resource units to FA_failure-affected MNs
   For each failure-free FA2
      If (lower_mark)—select the backup members from the FAs with lower traffic
         If(available_resources [i]>=a threshold)
           Call Be_Backup_Member
      Else--need to select the backup members from the FAs with high traffic
            Call Be_Backup_Member
   Lower_mark          false
 Until MN_assigned>=FA_failure_affected_MNs
 /*specify a failure_free FA to be a backup member */
 Subroutine Be_Backup_Member
    resources _denoted r x available_resources[i]
    /* r is the ratio donating available resources for FA_failure-affected MNs*/
    If (resource_donated>)
        Specify the FA2 as a backup member
        available_resources[i] available_resources[i]-resources_donated
        resources_to_use[i] resources_to_use[i] + resources_donated
        MNs_assigned MNs_assigned + resources_donated

Fig 4b. Selecting the backup members         MNs to the serving areas of the backup
The third task, as shown in Fig.4c,          members and to update the mobility
performs the system-initiated handoff        bindings of such MNs. From section 3.1,
both to virtually move FA-                   we know that the virtual moves of the
FA_failure-affected MNs are achieved            backup members is specially designated
by modifying the FA-serving records of          as the HA’s backup manager.
the failure-affected RANs. However the          The second task restores the mobility
resources     of     backup   members     are   bindings of the faulty HA by sending a
derived from a portion of available             mobility-reconstruction message to each
resources in a failure-free FA. It is           FA. Upon receiving message each FA
possible that the MNs in the failure-           searches its visitor list to find the visitors
affected RAN cannot be fully served by          entry with the identifier of the faulty FA.
a single backup member. Each failure-           Each qualified visitor entries is then
affected RAN first needs to specify             reorganized in the form of a mobility
enough back up members to serve the             binding entry and sends back to the
MNs in its coverage area. Then it resets        HA’s    backup     manager.     The    HA’s
its FA-serving record to the identifiers of     backup manager divides its received
specified backup members. Next each             mobility binding entries into several
specified backup member adds the                groups based on their corresponding
visitor    entries    for   the   FA_failure-   MN’s IP addresses and then sends each
affected MNs served by it. Meanwhile            group to a backup member. If a backup
the HAs of the FA_failure-affected MNs          member is designated to maintain the
are notified to update their mobility           mobility bindings of MNs with the IP
bindings.                                       address in the range [IP1,IP2],it will take
                                                the responsibility for intercepting the
4.2 IMPLEMENTATION OF HOME                      packets whose destination addresses
AGENT                                           belong to [IP1,IP2],and then tunneling
          There are three main tasks in the     such packets.
procedure for implementing the fault                     The third task is to change the
tolerance of the HA. Based on HA                packet interceptors of the HA failure-
loading status given by the OA&M, the           affected MNs from the faulty HA to the
first task selects the multiple failure-free    backup members. To change the packet
HAs with low traffic to be the backup           interceptors, the collocated routers of the
members of the faulty HA. One of the            faulty HA must remove their routing
                                                entries corresponding to the faulty HA
and     add      the     routing      entries    respective HAs of the failure-affected
corresponding to the backup members.             MNs are also notified to update the
However, the backup members may not              MN’s mobility bindings.
have direct physical interfaces with the
collocated routers. In each routing entry
corresponding to a backup member, its             When a faulty HA is recovered, the
interface field is set to a virtual interface.   responsibility of packet interception can
The virtual interface points to a software       be returned to it. By modifying the
program to perform the packet tunneling.         routing tables of the collocated routers of
By tunneling the packets destined to a           the   recovered      HA,     the    packet
HA-failure-affected MN are intercepted           interceptors of HA-failure affected MNs
by backup members (HA1 - HA4), not               are changed back from the backup
the faulty HA (HA0).                             members to the recovered HA. In
                                                 addition, the mobility bindings of the
                                                 recovered HA are also required to be
4.3 FAILURE RECOVERY                             reconstructed, which can be done by
                                                 searching all FA’s visitor lists. The
When a faulty FA is recovered from               reconstruction task is similar to the task
failure, the FA-failure-affected-MNs can         of restoring the mobility bindings of the
be served back by the recovered FA.              faulty HA.
First, the recovered FA determines the
information which RANs are preserved
by it (the information about the failure-
affected RANs). Then, the failure-
affected RANs reset their FA-serving
records to the identifier of the recovered
FA. The recovered FA gets back be the
common serving FA of all FA-failure-
affected MNs. Next, the recovered FA
creates a lot of visitor entries for the
failure-affected MNs. In addition, the
           This paper has presented an
efficient approach to tolerating the
failures of mobility agents in a wireless
system. The proposed approach utilizes
the available resources in other failure-
free mobility agents to dynamically
generate a backup set for each faulty
mobility     agent.   Compared        to   the
previous     approaches,     the     proposed
approach      possesses      the    following
      Not requiring hardware support.
      Not       incurring         failure-free
      Disturbing      the     fault-tolerant
       overhead to avoid the significant
       performance degradation on a
       single failure-free mobility agent.

Shared By: