1. © 2009 IBM Corporation Unlocking Wireless Performance with Co- operation in Base-Station Pools Parul Gupta, IBM Research – India COMSNETS - Jan 8, 2010

2. © 2009 IBM Corporation2 Overview  Why Co-operate?  Base Station co-operation in present network architecture  Pooled Base Station architecture  Potential cost savings through pooled BS model for a few scenarios –Interference Avoidance –Interference Alignment –Uplink Macro-Diversity –Efficient handovers  Summary and Future work

3. © 2009 IBM Corporation3 Why Co-operate?  There is demand for supporting many users with high data rates at high mobility. Challenges: –Spectrum is limited: Reuse desirable –For systems with spectrum reuse, capacity is fundamentally limited by interference –With the trend towards smaller cells for reducing transmit power and better reuse, handovers become more frequent  Base Stations (BS) can co-operate to –Spatially multiplex many independent data streams on the same channel. Prior work shows increased channel rank for such virtual arrays [1] –Distributed Transmit Beamforming –Interference Avoidance and Interference Cancellation –Load Balancing via joint-scheduling –Reduces latency during handoff, necessary for real-time applications like VoIP and streaming video [1] V. Jungnickel, S. Jaeckel, L. Thiele, L. Jiang, U. Krger, A. Brylka and C.V. Helmolt, “Capacity measurements in a cooperative MIMO network”,IEEE Transactions on Vehicular Technology, vol. 58, no. 5, pp. 2392-2405, Jun 2009.

4. © 2009 IBM Corporation4 Co-operation in Distributed Network Architecture  Assumption of infinite backhaul not always true –US has 75% copper, 15% fiber and 10% microwave. –Companies like Clearwire are leasing T1 bundles for their new network deployment: 6 T1s per Wimax BS in Manhattan! –Cost increases with each extra T1-line leased: $400 p.m. for 1.54 Mbps  Some co-operation schemes might still be possible in the distributed network architecture with limited backhaul  Schemes need to be designed appropriately for constraints, e.g. limited co-operation  There is a cost associated with communication over the backhaul: whether over a peer-peer BS interface (where exists) or a higher hierarchical element like RNC or ASN Gateway

5. © 2009 IBM Corporation5 BS Radio network controller Mobile switch center Service support node Gateway PSTN Access NetworkCore Network Present 2G-3G Wireless Network architecture Service Network SMS/MMS WAP GW 4G Wireless Network with Co-located Base-Station Pools Internet SMS/MMS IMS Content Service Web Service BS cluster Edge gateway Management Server Billing Edge gateway

6. © 2009 IBM Corporation6 Base Station Pools eliminate communication costs in co-operation  Information resides in a common place, transparently accessible to all BSs  Make fine-grained communication possible  Co-operation schemes require exchanging high volumes of data in short times become realizable  In this work, we estimate the potential cost savings for a few such schemes

7. © 2009 IBM Corporation7 Interference Avoidance  Capacity of full frequency reuse systems gets limited due to interference, esp. for cell-edge users  Interference can be avoided with joint resource allocation and power control, e.g. Fractional Frequency Reuse  Less complex, but takes a capacity hit  Each BS needs to share its power information with neighbors Cell 1 Cell 2 Cell 3 Full Frequency Reuse System Fractional Frequency Reuse System

8. © 2009 IBM Corporation8 Interference Avoidance – Example Communication Cost  Relative Narrowband Transmit Power (RNTP) messages specified in LTE specifications can indicate interference in the Downlink  Contain a bitmap for each Resource Block (100 per slot in 20 MHz bandwidth)  Similarly for Uplink, Interference indicator messages restricted to once every 20 ms to avoid excess overhead

9. © 2009 IBM Corporation9 Interference Cancelation  Dimensionality of channel matrix with K transmitters and receivers: K 2  For sharing this information with all co-operating BSs, communication cost grows as K 3  Example backhaul calculations are done assuming the complex CSI for the 720 data subcarriers, 10 MHz Wimax channel, fed back every 10 ms  Note: Spectrum to feedback CSI to the transmitter potentially an issue. TDD systems can utilize channel reciprocity to estimate downlink-CSI [1] V. Cadambe and S. A. Jaffer, “Interference alignment and degrees of freedom for the K-users interference channel,” IEEE Transactions on Information Theory, vol. 54, no. 8, pp. 3425-3441, Aug 2008 [2] H. Zhang et. Al., “Asynchronous Interference Mitigation in Co-operative Base-Station Systems”, IEEE Transactions on Wireless Communications, Vol. 7, No. 1, Jan 2008  Rather than avoiding interference, co-operating BSs can pre-code the transmitted signals to minimize interference at the receiver –Interference alignment [1] –Asynchronous Interference mitigation [2]  More complex because of signal processing  Assumes all co-operating BSs have full Channel State Information (CSI) at the transmitter

10. © 2009 IBM Corporation10 Uplink Macro-Diversity  Macro-Diversity schemes today (e.g. in Macro- Diversity Handover in Wimax) in the uplink rely on selection diversity  The extra gains due to Maximal Ratio Combining are untapped due to large amounts of data exchange and computation complexity  Example calculation shown for communication cost for 10 MHz Wimax channel, 2:1 DL:UL ratio, 5 ms frame, assuming 3 samples need to be transmitted per subcarrier  The amount of data to be transferred over the network is large, even for few quantization bits  Base-Station Pools eliminate this communication cost over the network, making MRC realizable h1h1 h2h2 x y1y1 y2y2

11. © 2009 IBM Corporation MS Serving BS (#1) Target BS (#2) Target BS (#3) MOB_NBR-ADV MOB_MSHO-REQ BS #2, BS #3 MOB_BSHO-RSP Handover to BS # 2 MOB_HO-IND DL/UL MAP, DCD/UCD RNG-REQ AUTHENTICATION Resume normal operation MOB_SCN-REQ MOB_SCN-RSP RNG-RSP Multiple iterations to adjust local parameters REG-REQ REG-RSP Service interruption duration … RNG-REQ RNG-RSP End Tx/Rx Scan Channel RNG_REQ MOB_ASC_REPORT RNG_RSP CONTEXT TRANSFER Shorter ranging cycle Resume normal operation Faster Handovers with Co-operation

12. © 2009 IBM Corporation12 Faster Handovers with Co-operation  Handovers can be made faster by –Co-ordination between base stations for ranging –Transfer of static context (service flow, authentication & registration info) and dynamic context (ARQ states, pending data) BS1BS2BS3 Shared MS data Co-located Base Station Pool

13. © 2009 IBM Corporation13 Summary and Future Work  Co-operation between Base Stations can improve wireless system performance in various ways –Interference Avoidance and Interference Cancellation –Load Balancing via joint-scheduling –Macro-Diversity Schemes –Faster Handovers  Fine-grained co-operation becomes possible due to transparent information sharing in Base- Station Pools  So far, we have set the motivation for co-operation in BS pools through estimating potential cost-savings. Future work would be to demonstrate working schemes in a BS pool and solve associated issues.