You are on page 1of 53

A

Close Examina.on of Performance and Power Characteris.cs of 4G LTE Networks


Junxian Huang1 Feng Qian1 Alexandre Gerber2 Z. Morley Mao1 Subhabrata Sen2 Oliver Spatscheck2
1University of Michigan

2AT&T Labs - Research

June 27 2012

LTE is new, requires explora.on


4G LTE (Long Term Evolu/on) is future trend
Ini.ated by 3GPP in 2004
100Mbps DL, 50Mbps UL, <5ms latency

Entered commercial markets in 2009

Lessons from 3G UMTS networks


Radio Resource Control (RRC) state machine is important App trac pa\erns trigger state transi.ons, dierent states determine UE power usage and user experience State transi.ons incur energy, delay, signaling overhead

LTE state machine

LTE power model

Network performance Energy eciency Parameter congura.on Mobile applica.on

RRC state transi.ons in LTE


Continuous Reception

Ti Ttail
Short DRX

DRX Long DRX

Tis

RRC_CONNECTED
Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


No radio resource allocated Ti Low power state: 11.36mW average power Tis
Short Long DRX Promo.on delay from DRX Continuous RRC_IDLE Reception

DRX

Ttail

RRC_IDLE to RRC_CONNECTED: 260ms RRC_CONNECTED


Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


Continuous Reception

RRC_CONNECTED

Ti
Short DRX

Tis

Long DRX

Radio resource allocated DRX Power state is a func.on of Ttail data rate: 1060mW is the base power consump.on Up to 3300mW transmicng at full speed

RRC_CONNECTED
Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


Con/nuous Continuous Send/ Reception Recep/on Prom receive a p ote to RRC_C acket
Ti
Short DRX

ONNE CTED

DRX

Ttail Tis
Long DRX

Reset Ttail
RRC_CONNECTED
Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


Continuous Reception DRX DRX Long DRX

Ti
Short DRX

Tis

x e l i a Tt

Ttail es pir

RRC_CONNECTED
Timer expiration

Ttail stops Demote to RRC_IDLE RRC_IDLE


Data transfer

Tradeos of Ttail secngs

Ttail seIng Long Short

Energy Consump/on High Low

# of state transi/ons Small Large

Responsiveness Fast Slow

RRC state transi.ons in LTE


Continuous Reception

DRX T i Listens to downlink channel periodically for a short Ttail dura.on and sleeps for the rest .me to save energy Tisof responsiveness at the cost Short Long
DRX DRX

DRX: Discon/nuous Recep/on

RRC_CONNECTED
Timer expiration

RRC_IDLE
Data transfer

Discon.nuous Recep.on (DRX): micro-sleeps for energy saving


In LTE 4G, DRX makes UE micro-sleep periodically in the RRC_CONNECTED state
Short DRX Long DRX

DRX incurs tradeos between energy usage and latency


Short DRX sleep less and respond faster Long DRX sleep more and respond slower

In contrast, in UMTS 3G, UE is always listening to the downlink control channel in the data transmission states

A DRX cycle consists of

DRX in LTE

On Dura.on - UE monitors the downlink control channel (PDCCH) O Dura.on - skip recep.on of downlink channel

Ti: Con.nuous recep.on inac.vity .mer


When to start Short DRX

Tis: Short DRX inac.vity .mer


When to start Long DRX Data transfer Ti expires On Duration Ti starts Tis starts

Tis expires

Long DRX cycle

Continuous Reception Short


DRX cycle

Long DRX cycle

LTE state machine

LTE power model

Network performance Energy eciency Parameter congura.on Mobile applica.on

Power trace of RRC state transi.ons


4000 Power (mW) 3000 2000 1000
t1: Promotion starts t4: Tail ends t3: Tail starts t2: Data transfer starts

0 t1t2

10 t3 15 Time (second)

20 t4

25

The data points are sampled and DRX in RRC_CONNECTED tail is not obvious due to the low sampling rate

Measured with a LTE phone and Monsoon power meter, averaged with repeated samples

LTE power model

Measured with a LTE phone and Monsoon power meter, averaged with repeated samples

LTE power model

Measured with a LTE phone and Monsoon power meter, averaged with repeated samples

LTE power model

Measured with a LTE phone and Monsoon power meter, averaged with repeated samples

LTE power model

Measured with a LTE phone and Monsoon power meter, averaged with repeated samples P(on) P(o) = 620mW, DRX saves 36% energy in RRC_CONNECTED High power levels in both On and O dura/ons in the DRX cycle of RRC_CONNECTED

LTE power model

LTE consumes more instant power than 3G/WiFi in the high-power tail
Average power for WiFi tail
120 mW

Average power for 3G tail


800 mW

Average power for LTE tail


1080 mW

Power model for data transfer


A linear model is used to quan.fy instant power level:
Downlink throughput td Mbps Uplink throughput tu Mbps

< 6% error rate in evalua/ons with real applica/ons

Energy per bit comparison


LTEs high throughput compensates for the promo.on energy and tail energy Transfer LTE WiFi 3G Size J / bit J / bit J / bit 10KB 10MB 170 0.3 6 0.1 100 4

Total energy per bit for downlink bulk data transfer

Energy per bit comparison


LTEs high throughput compensates for the promo.on energy and tail energy Transfer LTE WiFi 3G Small data transfer, LTE wastes energy Size J / bit J / bit J / bit Large data transfer, LTE is energy ecient 10KB 170 6 100 10MB 0.3 0.1 4
Total energy per bit for downlink bulk data transfer

LTE state machine

LTE power model

Network performance Energy eciency Parameter congura.on Mobile applica.on

4GTest on Android

Network characteris.cs

h>p://mobiperf.com/4g.html Measures network performance with the help of 46 M-Lab nodes across the world 3,300 users and 14,000 runs in 2 months 10/15/2011 ~ 12/15/2011
50 45 Latitude 40 35 30 WiFi 25 WiMAX LTE 20 -130 -120

-110

-100 -90 Longitude

-80

-70

4GTest user coverage in the U.S.

Downlink throughput
LTE median is 13Mbps, up to 30Mbps
The LTE network is rela.vely unloaded

WiFi, WiMAX < 5Mbps median


30 Y1: Network throughput (Mbps) 25 20 15 10 5 0 WiFi LTE WiMAX eHRPD EVDO_A 1

Uplink throughput
LTE median is 5.6Mbps, up to 20Mbps WiFi, WiMAX < 2Mbps median
30 Y1: Network throughput (Mbps) 25 20 15 10 5 0 WiFi LTE WiMAX eHRPD EVDO_A 1

RTT
LTE median 70ms WiFi similar to LTE WiMAX higher
30 Y1: Network throughput (Mbps) 25 20 15 10 5 0 WiFi LTE WiMAX eHRPD EVDO_A 1

LTE state machine

LTE power model

Network performance Energy eciency Parameter congura.on Mobile applica.on

User trace based analysis


UMICH data set
Collected from 20 volunteer smartphone users for ve months, totaling 118GB Contains packet traces including full payload

Trace-driven modeling methodology


Network model simulator
Simulates network states, such as RRC state transi.ons

Power model simulator


Calculates power usage based on the network states

Total energy usage


30 25 20 15 10 5 0

Comparing total energy of all user traces via simula.on in LTE/3G/WiFi


LTE/WiFi 23 3G/WiFi 15
Energy ratio: LTE/WiFi Energy ratio: 3G/WiFi

Energy ratio

All

3 User ID

Energy consump.on break down


Tail energy dominates LTE energy consump.on, similar to 3G
% of total energy 100 80 60 40 20 0 All 1 2 3 User ID 4 5 6 % Idle energy % Tail energy % Promotion energy % Data transfer energy LTE WiFi 3G

The total energy for dierent networks and users is normalized to be 100%

Energy consump.on break down


Tail energy dominates LTE energy consump.on, similar to 3G
% of total energy 100 The tail problem is the key factor for 80 LTEs high energy consump.on, 60 similar to 3G networks 40 20 0 All 1 2 3 User ID 4 5 6 % Idle energy % Tail energy % Promotion energy % Data transfer energy LTE WiFi 3G

The total energy for dierent networks and users is normalized to be 100%

LTE state machine

LTE power model

Network performance Energy eciency Parameter congura.on Mobile applica.on

Impact of conguring LTE tail .mer (Ttail)


S is dened to be the number of promo.ons Ttail has signicant impact on radio energy E, channel scheduling delay D, and signaling overhead S
(relative change) 1.5 1 0.5 0 -0.5 0 5 TD 15 20 Ttail (second) 25 30 E D S

TD is the default secng for Ttail in the measured network

LTE state machine

LTE power model

Network performance Energy eciency Parameter congura.on Mobile applica.on

App case study


Studied 5 web-based apps LTE has comparable page loading .me as WiFi, with 3G lagging behind CPU usage for LTE/WiFi is between 80% ~ 90% during page loading
Network does not appear to be the bo\leneck

Total energy consump.on: LTE > 3G >> WiFi

App case study


Studied 5 web-based apps LTE has comparable page loading .me as WiFi, In LTE network, applica.ons with should 3G lagging behind more aggressively burst CPU utrac sage for LTE/WiFi is between 80% ~ 90% to make m ore ecient during page loading use of t he bandwidth given the Total energy consump.on: LTE > 3G >> WiFi
Network does n appear to be the bo\leneck high eot nergy overhead

Summary
LTE has signicantly higher speed, compared to 3G and WiFi LTE is much less power ecient than WiFi due to its tail energy for small data transfers Derived a power model of a commercial LTE network, with less than 6% error rate UE processing is the bo\leneck for web-based applica.ons in LTE networks Mobile app design should be LTE friendly

Thank you!
Q & A
Contact: Junxian Huang (hjx@umich.edu)

Backup slides

Power trace of DRX in RRC_CONNECTED


2000 1800

Power (mW)

1600 1400 1200 1000 800 0 50 100 Time (ms) 150 200

RRC state transi.ons in LTE


Send/ receive a packet Con/nuous Continuous Reception Recep/on
Ti Ttail
Short DRX DRX Long DRX

Tis

Reset Ti RRC_CONNECTED
Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


Continuous Reception

Ti Ttail

DRX Long DRX

Short Short DRX DRX

Tis

Ti stops, Tis starts


RRC_CONNECTED
Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


Con/nuous Continuous Reception Recep/on
Ti Ttail
Short DRX DRX Long DRX

Tis

Reset Ti, stops Tis RRC_CONNECTED


Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


Continuous Reception

Ti Ttail
Short DRX

DRX

Tis Tis expires

Long Long DRX DRX

Tis stops RRC_CONNECTED


Timer expiration

RRC_IDLE
Data transfer

RRC state transi.ons in LTE


Con/nuous Continuous Reception Recep/on
Ti Ttail
Short DRX DRX Long DRX

Tis

Reset Ti RRC_CONNECTED
Timer expiration

RRC_IDLE
Data transfer

Impact of DRX inac.vity .mer (Ti): Con.nuous recep.on to short DRX


Dierently, S is dened as the sum of the con.nuous recep.on .me and DRX on dura.ons in RRC_CONNECTED Ti has negligible impact on E, however, S is signicantly aected
2.5 2 1.5 1 0.5 0 -0.5
(relative change)

E D S

0 TD 200

400 600 Ti (ms)

800

1000

Interes.ng ques.ons about LTE


To users: what is the end performance?
Network performance, such as RTT and throughput, how it compares with WiFi, 3G and WiMAX, etc. Energy eciency aec.ng ba\ery life, is LTE more power ecient than 3G or WiFi?

To ISPs: what is the impact of conguring LTE- related parameters on UE power saving, and delay/signaling overhead? To OS/applica.on developers: what is the performance bo\leneck of applica.ons in LTE network, CPU or network speed?

Energy per bit comparison


For large data transfer with maximum rate, LTEs energy eciency is comparable with WiFi, due to LTEs high downlink throughput
180 160 140 120 100 80 60 40 20 0 10 LTE DOWN LTE UP WiFi DOWN WiFi UP 3G DOWN 3G UP

J / bit

100 1000 Bulk data size (kB)

10000

One way delay and impact of packet size (not quite related)
LTE uplink one way delay (OWD) is larger than that of downlink RTT in LTE is more sensi.ve to packet size than WiFi, mainly due to uplink OWD
100 80

Delay (ms)

60 40 20 0 0 UP OWD DOWN OWD RTT 200 400 600 800 1000 1200 Packet size without TCP/IP headers (byte) 1400

JavaScript execu.on speed: a representa.ve view of smartphone processing capability


From 2009 to 2011, smartphones have signicantly improved JavaScript execu.on speed
Laptop/Chrome Mac OS X 10.7.2 iPhone 4S/Safari iOS 5.0.1 iPhone 4/Safari iOS 5.0.1 HTC/Default Android 2.2.1 Samsung/IE Windows Phone 7.5 G1/Default Android 1.6 G1/Default Android 1.6 iPhone/Safari iOS 3.0 Samsung/IE Windows Mobile 6.5

0.41s 2.26s 3.93s 4.42s 9.46s

Tested in Nov, 2009 Tested in Nov, 2011

95.08s 95.66s 117.41s 210.97s

0 50 100 150 200 250 Time to finish JavaScript benchmark (sec)

Power model for data transfer


A linear model is used to quan.fy instant power level:
Uplink/downlink throughput tu/td (Mbps)

< 6% error rate for predic/ng energy usage of 5 real applica/ons