On Integrated Location and Service Management for Minimizing Network Cost in Personal Communication Systems (by I R Chen, Baoshan Gu and S-T Cheng) Presented by Arnold Boedihardjo and Vivek Srivastava December 1, 2004

2 Agenda Introduction Motivation Integrated service and location management schemes Cost models Performance evaluation Results Conclusions

3 Introduction Personal Communication Service (PCS) Wireless voice and/or data communication service Includes technologies such as GSM 1900 and CDMA Digital cellular service at 1900 MHz Example services Banking Stock quotes Travel information Weather … Devices and mediums Phones / cellular network Blackberry / cellular network …

4 PCS system: A simple view Client and server model with mobility Clients – Mobile devices Server Client

5 PCS system: Location management PCS signaling network architecture Home Location Register Public Switched Telephone Network Regional Signaling Transfer Point Local Signaling Transfer Point Visitor Location Register

6 PCS system: Problem Managing services in roaming environment How does the server find new VLR? Too much overhead in contacting HLR VLR Server

7 PCS system: Solution Service Proxy Proxy Client Server

8 PCS System: Solution Service Proxy Context Information Tracks the location of the MU Forwards client requests to servers Forwards server replies to clients Reduces communication costs —Eliminates server overhead to check with location management system before data delivery

9 Static service proxy Proxy at a fixed location Inefficient data route delivery Proxy Client Server

10 Mobile service proxy Proxy moves with client Proxy informs server of location changes Server Proxy/ Client Proxy/ Client Proxy/ Client

11 PCS decoupled model Service and location management are decoupled Problem: Extra communication cost incurred for updating service proxy HLR / VLR Client Proxy

12 Integrated location and service management Combine location and service handoffs An implementation proposed by Jain and Krishnakumar [8,10] Requires fully replicated servers Example: Video stream —Transfer of service context to new server —Service context –Time marker, title, current frames, etc Problem: Fully replicated servers is difficult to attain for large deployment

13 A better integrated solution Fully replicated servers is not a prerequisite Address general personalized services Per-user proxy service Proxy service colocated (moves) with location database Centralized, fully distributed, dynamic anchor and static anchor scheme

14 Centralized scheme Proxy is centralized and colocated with the HLR to minimize communication costs with HLR to track MU When MU moves to different VLR invokes a location update operation and no server update A call delivery invokes a search operation at the HLR to locate the MU Service route: MU proxy/HLR server

15 Fully distributed scheme Location and service handoffs occur when MU moves to new VLR Service proxy colocated (moves) with location database at the current VLR Moving to new VLR invokes location update, context transfer and server update Call delivery invokes search operation at the HLR to locate the MU Service route: MU/proxy server

16 Static anchor scheme VLR grouped into logical boundaries called anchor boundaries HLR now contains anchor information Proxy co-located with anchor and is fixed until MU is within the anchor area Intra-anchor movement Location update to anchor of new VLR with no update to the HLR Inter-anchor movement Location update to HLR of new anchor, service context transfer and update server Call delivery invokes a search operation at the HLR to locate the current anchor VLR and subsequently to the MU Service route MU proxy/anchor server

17 Dynamic Anchor Scheme Same as static anchor except that the anchor changes to current VLR within intra-anchor movement through a call delivery Call delivery Search operation to the HLR If local anchor is not current serving VLR —Location update, service context transfer, update server Service route: MU proxy/anchor server If there is a service request following a call delivery, current VLR is anchor VLR the service route is MU server which is a benefit over static anchor scheme

18 Variables used λthe average rate at which the MU is being called. σthe average rate at which the MU moves across VLR boundaries. γthe average rate at which the MU requests services. CMRcall to mobility ratio, e.g., λ / σ. SMRservice request to mobility ratio, e.g., γ / σ. Tthe average round trip communication cost between a VLR and the HLR (or between a VLR and the server) per message. the average round trip communication cost between the anchor and a VLR in the anchor area per message. the average round trip communication cost between two neighboring anchor areas per message. the average round trip communication cost between two neighboring VLRs per message. M cs the number of packets required to transfer the service context. NsNs the number of server applications concurrently engaged by the MU. P InA the probability that a MU moves within the same anchor area when a VLR boundary crossing movement occurs. P OutA the probability that a MU moves out of the current anchor area when a VLR

19 Cost models Performance metric – total communication cost per time unit 3 basic operations Location update ( ) – cost for updating the location of MU and service proxy (sometimes, service context transfer) Call delivery ( ) – cost for locating a MU to deliver a call User service requests ( ) – cost for MU to communicate with server through proxy

20 Centralized and fully distributed scheme costs Centralized scheme : cost to inform the location database at the HLR of the current VLR : cost to locate the MU and deliver the call from the HLR to VLR : round trip cost from MU to service proxy and proxy to server) Fully distributed scheme — : cost to inform the location database at the HLR about the current VLR — : cost to transfer the service context between two neighboring VLRs — : cost to update Ns application servers of the new location of MU : cost to locate the MU and deliver the call : cost from the proxy to the server

21 Centralized and fully distributed scheme costs Type of scheme/type of cost Location updateCall deliveryService request Centralized :cost to inform the location database at the HLR of the new VLR :cost to locate the MU and deliver the call from the HLR to VLR :round trip cost from MU to service proxy and proxy to server Distributed :cost to inform the HLR of the new VLR :cost to transfer service context between two neighboring VLRs :cost to update Ns application servers with new location of MU :cost to locate the MU and deliver the call :cost from the service proxy co- located at the current VLR to the server

22 Dynamic anchor scheme model (1) Call: Movement: CE Cs ServCvdC ServNonCvdC InA OutA InMs OutMs Service Request: ME Ms ServOutM ServInMFlag RstFlag Sink Flag RstFlag1 ServNonCvdS ServCvdS SE Ss RstFlag2 Note: Refer to the paper for meaning of symbols

23 Dynamic anchor scheme model (2) Call delivery Call arrives and Cs is filled with a token If mark(Flag) > 0 —ServNonCvdC is enabled and current VLR is not same as anchor VLR —HLR is queried to locate the anchor; anchor queries the current serving VLR and MU’s location is returned —Anchor is moved to current VLR If mark(Flag)=0 —ServCvdC is enabled and current VLR is same as anchor VLR —HLR is queried to locate the anchor; anchor returns the MU’s location immediately

24 Dynamic anchor scheme model (3) Location update When MU moves, token is placed in Ms If movement is intra-anchor with probability InA —ServInM is enabled and a local anchor update is performed —mark(Flag) > 0 indicating current VLR != anchor VLR If movement is inter-anchor with probability OutA —ServOutM is enabled and the HLR is updated with current VLR (new anchor) —service context is transferred from old anchor to new anchor and, —application servers are updated with new address of proxy —mark(Flag) = 0 indicating current VLR = anchor VLR

25 Dynamic anchor scheme (4) Service request MU sends a service request and token is placed in Ss If mark(Flag) > 0 —ServNonCvdS is enabled and current VLR is not anchor VLR. —Request is sent to service proxy colocated with local anchor to forward to server —Service proxy is co-located with anchor, so no extra cost to obtain MU’s current location If mark(Flag) = 0 —ServCvdS is enabled and current VLR is the anchor VLR —Service proxy is co-located with anchor VLR

26 Dynamic anchor scheme model (5) System costs N states in the underlying Markov model :steady state probability that system is in state i :search cost assigned to state i under the dynamic scheme :cost for location update in state i : cost for service request in state i

27 Dynamic anchor scheme model (6) System costs (cont’d) Expected search cost under the dynamic scheme Expected location update cost under the dynamic scheme Expected service request cost under the dynamic scheme

28 Static anchor scheme model (1) Cs CE Call: ServC Movement: InA OutA InMs OutMs ME Ms ServInM ServOutM Service Request: SESsServS

29 Static anchor scheme model (2) Anchor is fixed in an anchor area until the MU departs from that area Call delivery Cost from the HLR to the anchor (T) + anchor to the current VLR ( ) Location update If movement is intra-anchor with probability InA —ServInM is enabled and a local anchor update is performed If movement is inter-anchor with probability OutA —ServOutM is enabled and HLR is updated with new anchor information —Service context is transferred from old to new anchor —Application servers are informed of the new anchor

30 Static anchor scheme model (3) Service request MU sends a service request and a token in placed in Ss ServS is enabled and service request is routed from the MU, to the proxy co-located with the anchor VLR and to the server System costs per unit time Call delivery: Location update: Service request:

31 Performance evaluation – Parameterization (1) Assumptions and basics Assume a hexagonal shaped cell n-layer VLR covers cells where n = 2 or n = 3 n-level LSTP contains VLRs n-level RSTP contains LSTPs One LSTP corresponds to one anchor area Probability that a MU moves within same anchor area Probability of an inter-anchor movement: Probability that a MU moves within the same RSTP e.g. n=2, boundary edges = 18, internal edges = 24. VLR

32 Performance evaluation – Parameterization (2) Communication between HLR and VLR moves through VLR-LSTP- RSTP-PSTN path Communication between two neighboring VLRs ( ): Communication cost between two VLRs in same LSTP( ) Communication cost between two VLRs in different LSTPs ( ) Communication cost between two VLRs out of same RSTP (1- ) Dynamic anchor scheme Communication between anchor VLR and another VLR in same LSTP:

33 Performance evaluation – Parameterization (3) Dynamic anchor scheme Communication between two neighboring LSTP areas ( ) —Communication between VLRs in same RSTP —Communication between VLRs in different RSTP Static anchor scheme Communication between anchor VLR and any VLR in the same LSTP (assuming MU has equal probability of being in each of the 7 VLRs) Communication between two neighboring LSTP areas –same as dynamic anchor scheme

34 Performance evaluation-Results (1) PCS network consists of 2-layer VLRs, LSTPs, RSTPs and HLR Normalized cost:,, Number of application servers: Call to Mobility ratio (CMR: ) – ratio of call arrival rate to mobility rate Service to Mobility ratio (SMR: ) – ratio of service request rate to mobility rate Cost rate – Total cost incurred per second (normalized with respect to the cost of transmitting a message between VLR and its LSTP)

35 Performance evaluation-Results (2) Cost rate under different CMR and SMR values

36 Performance evaluation-Results (3) Mobility rate fixed at 10 changes /hour SMR = 1 to study the effect of varying CMR Low call arrival – Centralized and fully distributed schemes worse than the anchor schemes High call arrival – Centralized scheme is best. Dynamic anchor scheme is better than static. Reason: Dynamic anchor scheme updates the HLR and moves the anchor to the current VLR thereby reducing service request costs and location update costs Cost rate under different CMR values

37 Performance evaluation-Results (4) Mobility rate fixed at 10 changes/hour CMR = 1 to study the effect of SMR on cost rate Low SMR – Fully distributed scheme is the worst due to movement of service proxy with mobility High SMR – Fully distributed scheme performs best since the service proxy is with the location database Cost rate under different SMR values

38 Performance evaluation – Results (5) Increase in context transfer cost leads to increase in cost rate Sensitivity: Fully distributed > Dynamic anchor scheme > Static anchor scheme > centralized Cost rate under different context transfer cost values

39 Performance evaluation – Results (6) Comparison of integrated with decoupled scheme Depending on the user’s SMR the “best” integrated scheme and decoupled scheme were compared Integrated scheme converges with the decoupled scheme at high SMR values where the “influence” of mobility is less Integrated scheme is better than basic scheme at high SMR values due to cost between the server and MU via HLR

40 Conclusions Concept: Position the service proxy along with the location database of the MU Centralized scheme: Suited for low SMR and high CMR Distributed scheme: Best at high SMR and high CMR Dynamic anchor scheme: Works best for a wide range of CMR and SMR values except when service context transfer costs are high Static anchor scheme: Works reasonably well for a wide range of CMR and SMR values