The Value of BS Flexibility for QoS-Aware Sleep Modes in Cellular Access Networks


Published on

Sleep modes are one of the most widely investigated techniques to decrease energy consumption in cellular access networks. However, the application of such algorithms on present day base stations (BS) equipment poses several challenges. Indeed, currently installed BSs are unfit for frequent on/off cycles. This may lead to increased failure rates and malfunctioning, ultimately resulting in significant CAPEX and OPEX increases for mobile network operators (MNOs). This situation calls for a new generation of flexible BSs endowed with a ”hot standby” mode, which guarantees quick activation times without affecting BS availability. However, when such new BS models become available, MNOs will need to determine a migration path to a new network deployment with progressive replacement of old BS equipment. In this paper, we propose an approach to quantify the benefits achievable by MNOs with the deployment of flexible BSs, in terms of maximum energy efficiency achievable with a given fraction of flexible BSs in their network. More specifically, we propose a method for estimating, for a given percentage of flexible BSs, the energy optimal density of static and flexible BSs, which is sufficient to serve a given set of active users with predefined performance guarantees. We show how to apply our method to derive bounds on the maximum energy savings achievable through sleep modes, as a function of the fraction of flexible BSs. We determine the effect of uncertainty in traffic predictions on sleep modes performance, and we derive indications for optimal network planning strategies.

Published in: Technology, Business
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

The Value of BS Flexibility for QoS-Aware Sleep Modes in Cellular Access Networks

  1. 1. The Value of BS Flexibility for QoS-Aware Sleep Modes in Cellular Access Networks Gianluca Rizzo HES SO Valais, Switzerland Balaji Rengarajan Accelera, USA Marco Ajmone Marsan Politecnico di Torino and Institute IMDEA Networks
  2. 2. What is the value of sleep modes for mobile network operators? • SM essential for achieving network energy proportionality – theoretical savings of up to 40-60% • Practical issues: – Legacy BS: slow dynamics – SM shorten their average lifetime – Flexible BS: Switching up speed vs standby energy tradeoff erodes gains – Uncertainty in traffic demand forecast • What is the optimal ratio of BS which should be replaced with flexible BS?
  3. 3. System Model Users, BS ~ Homogeneous PPP • Density is a function of time • Forecasted user density: Gaussian, with known mean Time is divided into periods and slots (within a period) Two types of base stations: • Static: They can only be on or off; • Flexible: They also access a low power standby state. A BS can transition from/to the off state only at the beginning of each period (slow sleep modes) The transition from (to) the standby state can happen at the beginning of each slot (fast sleep modes) Period: No off->on transitions Slots: Flexible BSs can sleep/wake up
  4. 4. A method for the derivation of the energy optimal BS density • QoS parameter: expected per-bit delay • The energy consumed by a BS with utilization U is • The energy optimal BS density is obtained by solving
  5. 5. Derivation of the optimal coordinated sleep modes strategy • We assume BSs are turned on/off independently – Sleep mode univocally identified by BS density • At each period: we compute the energy optimal total density of BS satisfying the QoS constraints, with a probability – For the peak user density, this corresponds to the max total n of BS – All BS in excess are turned off, starting from legacy BSs. • In each slot: we determine the optimal density of flexible BS, based on the value taken by user density in that slot – All flexible BS in excess are put into standby mode.
  6. 6. Numerical evaluation • Duration of each period: 3h – Max consumed power: 1500 W – Idle power consumption: 60% of max – Standby power consumption: 30% of max – .
  7. 7. Flexibility allows reducing the energy costs of uncertainty in traffic forecasts • Uncertainty affects also the total n of active BS
  8. 8. No prediction error • Fast sleep modes increase EE by at most 12% Combining slow and fast sleep modes enables the highest energy gains
  9. 9. • Need of an accurate evaluation of impact of slow sleep modes on capex and opex to determine the most efficient solution Energy cost of standby mode Combining slow and fast sleep modes enables the highest energy gains Only Flexible BSs (No BS is allowed to go off)
  10. 10. • Modest increase in energy efficiency • Capex/Opex of networks with slow sleep modes increase with frequency of power on/off events No need to speed up «slow» sleep modes
  11. 11. No need to speed up «slow» sleep modes • Majority of BS are turned off once per day: modest impact of on off costs 10% of idle power 30% of idle power
  12. 12. Conclusions • We provide a tool for estimating maximum achievable energy savings for a given % of flexible BS – Assuming a given QoS target is always met • We estimate the cost of uncertainty in traffic predictions • Our results enable CAPEX/OPEX analysis to determine optimal deployment strategies.
  13. 13. Future work Evaluate the impact of • different energy models • different traffic mix, with different requirements – Project with predicted evolution of wireless traffic Capex/Opex study • (include maintenance costs, projected energy costs, renewables, etc)
  14. 14. Thanks!