#### Transcript Celebi

Base Station Operation and User Association Mechanisms For Energy Delay Tradeoffs in Green Cellular Networks by Kyuho Son, Hongseok Kim, Yung Yi, Bhaskar Krishnamachari Haluk Celebi/ hc2489

Outline • • • • Motivation Problems to be solved System model Solution approaches and algorithms

Main Motivation • • BSs are large underutilized: Traffic is below 10% of peak 45% of the time.

UMTS BS consumes 800-1500W, and radio 30 40W. Thus, not only turning off the transceivers, we need to turn off BS completely.

### Problems to be solved

1. User association: When the set of active BS changes, MT needs to be associated with a new BS 2. Energy efficient BS operation: finding optimal set of active BS - Paper gives a distributed algorithm to solve 1 st problem optimally. For 2 nd problem, it offers several greedy algorithms.

• System Model and Problem Formulation arrival rate Traffic load density Set of active BS Transmission rate of user i located at x Fraction of time required to deliver traffic load from BS i to location x

Feasible solution to the problem: utilizations of BSs Note that in optimal case p(x) is either 1 or 0

Cost function • n is the parameter that determines between the flow level performance and the energy level consumption Flow level performance

is a parameter specifying the degree of load balancing Each MT prefers a BS that gives the highest rate Minimizing number of flows, thus delay in M/M/1 queue Thus as α increases, cost function places more importance on traffic load instead of transmission rate

### Cost function of Energy

Portion of fixed power consumption.

Energy that is spend only depends on the radio.

In other case, when q =1, regardless of the utilization amount of energy spend is Pi. That`s where turning BS off becomes more important We assume femto cells are more energy proportional (smaller q values) macro cells.

Solution approach to the problem • Solving two cost functions together is hard.

• We assume that flow arrival, and departure process is much faster than the period of determining active and passive BSs • Decompose problem into two sub-problems • Solve user association problem. For given active BS set find which BS each MT should connect to • BS operation problem

### Connection policy

It`s proved in the paper that with this policy, each MT is able to choose optimal BS