Optimal Dynamic Mobility Management for PCS Networks

Document Sample
Optimal Dynamic Mobility Management for PCS Networks Powered By Docstoc
					 Optimal Dynamic Mobility
Management for PCS Networks

        Li, Kameda, and Li
  IEEE Transactions on Networking
             June 2000

                                    1
              I. Introduction
• Mobile computing system
  – The integration of mobile communications and
    computing
• Personal communication system (PCS)
  – A new mobile computing system
    that enable users to economically transfer
    any form of information between any desired
    locations at any time
  – Cells
  – Base stations
  – Mobile switching center(MSC)                2
• Mobility management
  – How to track the mobile users that move from place to
    place in PCS networks
• Two basic operations in mobility management:
  – Location update
     • The process through which the system tracks the location of
       mobile users that are not in conversations
     • The up-to-date location information of a mobile user is reported
       by the mobile users dynamically
     • A location area may include one or more cells.
  – paging
     • When an incoming call arrives, the system searches for the
       mobile users by sending polling signal to cells in the location
       area.                                                          3
• Cost of location update and paging
  – Wireless bandwidth and processing power
    at the mobile users, base stations, and MSCs
  – A large location area will result in a decrease in the
    cost of location update and an increase in the cost of
    paging, and vice versa.
  – To determine the size of the location area is a critical
    problem for minimizing the total cost of location
    update and paging.


                                                         4
• Location update and paging schemes
  – Static
     • The location area size is fixed
     • Problem:
       excess location updates as a mobile user moves back and
       forth between two location areas
  – Dynamic
     • The location area size is determined dynamically according
       to the changes of mobility and calling patterns of mobile
       users


                                                             5
• Three dynamic location update schemes:
  – Distance-based
     • The location update is performed whenever the distance between the
       current cell for the mobile user and the last cell in which the last update
       was performed is d.
     • The location area is an area in which the central cell is the last cell
       where the last update occurred surrounded by d rings of cells
  – Movement-based                Ö (used in this paper)
     • The location update is performed whenever a mobile user completes d
       movements between cells (the location update movement threshold)
     • The location area is an area in which the central cell is the last cell
       where the last update occurred surrounded by d rings of cells
  – Time-based
     • The location update is performed every t units of time.
     • The size of location area is calculated according the the mobility of the
       mobile user user.
                                                                             6
II. Modeling and System Description
• The probability density function of
  cell residence time is pm(t) which has
  Laplace-Stieltjes transform fm(s) and mean 1/lm
• When a mobile user leaves a cell, there is an equal
  probability that any one of the immediate neighboring
  cells is selected as the destination.
• The movement-based location update scheme is
  considered in this paper.
   – A location update occurs when the number of boundary
     crossing since the last location update registration equals a
     threshold d.
   – The center cell is the cell where the last location registration
     occurs.
                                                                   7
• Assume that incoming call arrivals to each mobile
  user follow a Poisson process with rate lc
• As soon as a call for a mobile user arrives, the
  network initiates a paging process to locate the
  called mobile user.
• The paging area is the covering area within a
  distance d - 1 from the center cell.



                                               8
• Two popular cell configurations are studied in this paper:
   – Hexagonal cell configuration
   – Mesh cell configuration
     (Ri = the set of cells in the ith ring)




                                                        9
• The size of each cell is determined based on the number
  of mobile channels available per cell and the channel
  allocation scheme used.
• The location tracking mechanism can be applied to both
  macrocell (several kilometers) and microcell (several
  hundreds of meters) environments.
• The size and shape of cells are indirectly reflected by the
  cell residence time.
   – If the size of cells is small, the mean residence time will be
     relatively small, and vice versa.


                                                                  10
• Assume that homogeneous cells (of the same shape and
  the same size) are used.
• The distance is measured in terms of
  the number of rings such that
  the distance between a given center cell
  and the cells belonging to set Ri is i rings.
• The number of cells in ring i,



• Paging schemes:
  – Selective paging (Akyildix et al.)
     • The paging is performed in one of the subarea only
  – All cells in the location area are paged.               Ö   11
          III. Problem Formulation
• A. Cost of Location Update
  – U = the cost for performing a location update
  – a(j) = the probability that there are j boundary crossings
    between call arrivals
  – The expected location update cost per call arrival:



                                         d=1        d=2         d=3
    =U{     1 * [a(d) +…+ a(2d-1)]       1          2,3         3,4,5
          + 2 * [a(2d) +…+ a(3d-1)]      2          4,5         6,7,8
          + 3 * [a(3d) +…+ a(4d-1)]      3          6,7         9,10,11
          +… }                           … from 0   … from 0    … from 0
                                         1: to 1    1: to 2,3   1: to 3,4,5
                                         2: to 2    2: to 4,5   2: to 6,7,8
                                         3: to 3    3: to 6,7   3: to 9,10,11
                                                                     12
• 1) Calculation of a(j)
                           • tc = call interval
                           • R0 = cell at the last call
                           • tMi = period staying in
                             Ri
                           • tm = interval between
                             the last call and moving
                             out of R0
                           • tc,i = time between
                             entering Ri and the next
                             call

                                                13
• Let tMi (cell residence time) be an independent
  identically distribution (iid) random variable with
  – a general distribution function Gm(tMi),
  – density function gm(tMi),
  – and the Laplace-Stieltjes Transform:




                                                 14
• Let fc(t) and rm(t) be the density function of tc
  (call interval) and tm, respectively.
• Let E[tc] =1/lc and E[tMi] = 1/ lm
• Assume that the incoming phone call is a Poisson
  process:

• From the memoryless property of the exponential
  distribution, tc,i has the same exponential
  distribution as tc
                                               15
• for tm, from the random observer property
  [Stochastic Process - S. Ross], we have
  – The density function:
    (Eq. 5)
  – The Laplace-Stieltjes
    Transform:
    (Eq. 6)




                                              16
• The probability a(K) that user p moves across K
  cell’s between two phone calls is derived for
  K = 0 and K ³ 1:                 • where q = lc/lm,
  – For K = 0:                       the call-to-
                                     mobility ratio
                                     (CMR)
                                   • q < 1:
                                         lc < lm
                                         1/lc > 1/lm
                                         tc > tm
                                     tc call arrival time
                                     tm cell residence time
                                                    17
– For K ³ 1:




          Eq. 11   Eq. 9   Eq. 10



                                    18
• From (7) to (11), we have


                              19
• 2) Simplify
  the cost function Cu




                         20
• Substituting (13) into (2), we have




                   Eq. 15     Eq. 16


                                        21
22
• Substituting (15) and (16) into (14), we have




                                                  23
• B. Cost of Paging
  – Assume that the cost of polling a cell is P
  – The number of cells in a paging area with threshold
    value d:




  – The expected paging cost per call arrival:



                                                     24
• C. Total Cost per Call
  – TC(d)
  – The sum of the cost of location update, Cu, and the
    cost of paging , Cp:




                                                      25
    IV. Minimizing the Total Cost
• The goal of the optimal movement-based location
  update scheme
  – To find the optimal threshold, d,
    that minimize the total cost per call, TC(d)




                                                   26
• Theorem 1:
  – The location update cost, Cu(d), is a decreasing and
    convex function with respect to the threshold d;
    and the paging cost, Cp(d), is an increasing and
    convex function with respect to the threshold d.




                                                      27
– Proof:




           28
29
• Corollary 2: a direct result from Theorem 1
• Theorem 3:
  – The value of d is a unique solution to (21) if and only
    if the following relation holds.




  – That is,




                                                       30
V. Proposed Algorithm




                        31
• Step 1
  – Compares the values of –Cu’(d) and Cp’(d) at d = 1
  – where d > 0,
     Cu’(d) is decreasing and Cp’(d) is increasing
  – If Cu’(1) £ Cp’(1), the optimal threshold ð should be 1
• Step 2
  – Determines the interval [d,d+s] which consists of ð
  – ð should not be too large (e.g. > 20), thus we set ð = 10
• Step 3
  – Determines the the optimal threshold ð in the interval [d,d+1]
    by using binary search
• Step 4
  – Determines the the optimal threshold ð
                                                                32
        VI. Numerical Examination
• Assume that the call residence time follows the Gamma
  distribution.
• Let the Laplace-Stieltjes Transform , fm(s), of the
  Gamma distribution with mean 1/lm and variance n is


• We studies effects of optimal threshold d by various
  parameters:
   – Update cost U, polling cost P, CMR (Call-to-Mobility Ratio)
     q = lc/lm, and variance of the cell residence variance n
• Program in C and run on SPARC-20 workstation

                                                             33
• A. Effects of CMR, Update Cost, and Paging
  Cost




                                               34
35
• B.Effects of Cell Residence Time Variance




                                              36
37
38
39
40
41
42
43
          VII.Conclusion Remarks
• This paper studies a dynamic mobility management
  scheme: the movement-based location update scheme.
• An analytical model is applied to formulate the costs of
  location update and paging per call arrival.
• The problem of minimizing the total cost per call arrival
  is expressed as an optimization problem that finds the
  optimal threshold in the movement-based location
  update scheme
• An effective algorithm is proposed.


                                                       44

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:0
posted:7/14/2013
language:Unknown
pages:44