NII seminar - CityU CS - City University of Hong Kong

Document Sample
NII seminar - CityU CS - City University of Hong Kong Powered By Docstoc
					Minimal cost deployment of mesh
networks with QoS requirements
     in indoor environment

      Xiaohua Jia
      Dept of Computer Science
      City University of Hong Kong
Mesh Network Architecture
   Multihop
    WLAN (single hop)

   Gateway connection                Portal
                                       MP MPP
    MANET (no gateway)                             MP
                                  AP MAP
                                              MP        MP
                                              AP MAP    AP MAP
                                     Client                  Client
Mesh Network Planning
Problem: Given a set of users,                            Internet
   each with QoS requirements                    Portal
   (bandwidth and delay), find                    MP
   the optimal placement of AP,
   MP, and gateway nodes in the             MP
   area such that the users QoS             AP
   requirements are met and the                                 AP
   total cost of the AP, MP, and   Client
   gateway nodes is minimized.                   Client              Client
Output: 1) locations;
2) transmission power;
3) number of radios per AP.
Related Work
AP Placement in WLAN
[BCC07] S. Bosio, A. Capone, and M. Cesana, “Radio Planning of Wireless Local Area
   Networks,” IEEE/ACM Trans on Networking, vol. 15, no. 6, pp.1414 –1427, Dec 2007.
1) Min-set cover: place Min # of APs, such each client is covered by at least one AP;
2) Min overlap problem (MoP) / Max efficiency plan (MeP): given N of APs (or budget),
   place them such MoP or MeP is optimized.

[EGS07] A. Eisenblatter, H-F Geerdes and I Siomina, “Integrated Access Point Placement and
    Channel Assignment for Wireless LANs in an Indoor Office Environment”, IEEE Symp.
    on World of Wireless, Mobile and Multimedia Networks (WoWMoM), 2007.
1) Max avg throughput of all users for placing N APs. Each user’s throughput is f(dv,AP) under
    fixed power of APs;
2) Min overlap APs (in terms of number of clients) using the same channel;
3) LP formulation and computed by using CPLEX.

No multi-radio and rate adaption & power control.
Related Work (cont’d)
AP&MP Placement
[SL06] A. So and B. Liang, “Optimal Placement of Relay Infrastructure in Heterogeneous
      Wireless Mesh Networks by Bender’s Decomposition,” QShine’06.
1) place min # of relays in N users positions (served & connected);
2) Mathematical Programming formulation.

[WXC07] J. Wang, B. Xie, K. Cai, and D. Agrawal,
“Efficient Mesh Router Placement in Wireless
Mesh Networks”, IEEE MASS’07.
1) place min # MR among N candidate sites,
cover service area and interconnect relay nodes.
2) two steps: a) coverage; b) connectivity

No interference was considered.
Related Work (cont’d)
QoS Gateway Placement
[B04] Y. Bejerano, “Efficient Integration of Multihop Wireless and Wired Networks with QoS
    Constraints”, IEEE/ACM Trans on Networking, Vol. 12, No. 6, Dec 2004.
1) Transformed to: clustering of graph into min number of clusters; 2) QoS: cluster size and radius;
3) TDMA for intra-cluster and use of orthogonal channels for neighbor clusters.

[ABI06] B. Aoun, R. Boutaba, Y. Iraqi, and G. Kenward, “Gateway placement optimization in
    wireless mesh networks with QoS constraints,” IEEE Journal on Selected Areas in
    Communications, vol. 24, no. 11, pp. 2127 – 2136, Nov. 2006.
1) Graph partitioning based on k-hop Dominating-Set

No consideration of interference
for link capacity / throughput
Unique challenges
   Placement of different types of mesh nodes
    (AP, MP and gateway) and aiming at
    minimizing the total cost.

   APs can be equipped with different number of
    access radios.

   Each node (AP or MP) can adjust its
    transmission power and data rate depends
    on transmission power.
Decomposition of the problem

Subproblem 1: Optimal placement of APs to
 serve all clients.

Subproblem 2: Configure minimal number
 of Gateway nodes for a large cluster
 under QoS constraint.

Subproblem 3: Merge small clusters by
 adding minimal number of MPs
 AP Placement with multi-
 radios and power control
Problem: given a set of clients in an area, each client has
  bandwidth requirement γ. Place a set of APs W,
  determine number of radios for each node, and adjust
  power to meet γ, and the total cost is minimized:
    cos t (W )  p B | W |  p R  |  ( w) |

AP placement in indoor
   Divide the region into
   Traffic demands
    (Clients) originate
    from grids;
   APs are placed at
    the center of grids.
Transmission power, data rate
and interference
   AM×M: signal attenuation array
   Node v can receive data from w if:
    A(w,v)Pw ≥ α
   Node v can be interfered by w if:
    A(w,v)Pw ≥ β
   Data rate from v to w is (similarly for R(w,v)):
    R(v,w) = f(A(v,w)Pv)
A table of transmission range,
data rate and interference range
Interference model
Node interference

I(u) = {v| A(w,v)Pw ≥ β}                    u

Link interference
link l’ is interfered by l if one of
   the end-node of l’ is in the
   interference range of l.                     u   v
Interference and Bandwidth
Network G(V, E): V set of clients and APs. A link l in E
   is between a client and an AP.
I(l): collision set of link l is a set of links that either
   interfere with l or are interfered by l, including l itself.
Bandwidth constraint under
    vup ,  vdn: up link and down link traffics of v

   Channel bandwidth is shared by all links in the
    collision set I(l). That is:
                                           up
                                                         dn
    TI (l )        
                ( v , w)I ( l )
                                   (        v
                                       R(v, w)
                                                         v
                                                     R( w, v)
                                                                ) 1
A performance metric for
greedy algorithms
   S(w): clients served by AP w
   Max collision load:

   Client to Interference Ratio CIR(w):
Top-down method
1)        Initialization. Each client is placed with an AP.
2)        Choose two neighboring APs to merge to a new AP
          w, such that:
     a)    AP w can serve all clients of two old APs (w’s power is set
           to cover all clients);
     b)    CIR(w) is maximal (locate w’s new location);
     c)    Determine the number of radios according to S(w).
3)        Repeat step (2) until no more merge can be done.
Merging with neighboring APs
   The merge of APs should be between
    neighboring APs
   We use Delaunay graph of APs to ensure the
    merge between neighboring APs

Bottom-up method
1)   Initially all clients are not served.
2)   Place an AP at a grid and adjust it power such that
     the bandwidth constraint is met and:
      CIR(w) is maximal.
3)   Repeat the above step until all clients are served.
Simulation results
   100m×100m region divided into 20×20 grids
   pR: pB = 0.4 : 1 and γup : γdn = 1 : 9
Simulation results (Cont’d)
On-going research problems…
   Capacity analysis of using multiple access radios
    against use of single radio. What is the performance
    gain compared with the cost.

   k-coverage (k = 2) AP placement. Given per client’s
    bandwidth requirement γ1 if served by its primary AP,
    and γ2 if its primary AP failed, place minimal
    number of APs (and adjust power) such that each
    client is covered by at least k APs and γ1 and γ2 are
    met for fault tolerance.
Merging small clusters by
adding relay nodes (MPs)
   This problem is
    different from
    traditional minimum
    relay node
    placement problem
   The cost saving is
    the number
    gateways saved
    minus additional
    relay nodes
   QoS mesh network topology control problem
    can be divided into three subproblems and
    tackled separately
   The interference is the key for the
    performance of AP placement
   An effective metric for greedy algorithms can
    improve the performance significantly

Shared By: