KPIs for Energy Efficiency Network measurements - ETSI · PDF fileKPIs for Energy Efficiency Network measurements ... vs. C-RAN deployment etc. Energy Efficiency KPIs for ... QoE KPIs

KPIs for Energy Efficiency Network measurementsNetwork measurements

Erik Friman, [email protected]

1st ETSI TC EE Workshop

20-21 June 2012 , Genoa Italy KPI – Key Performance Indicator

Outline

› Introduction

› Drivers for Quality of Experience KPIs

– Traffic and application evolution

› QoS and QoE measurement scenarios

Energy Efficiency KPIs for network measurements | ETSI WS Genova 2012 | 2012-06-20 | Page 2

– Standardization QoE KPI for Video on demand

› Conclusion

IntroductionWhy measure EE on networks?

› Valuable for both mobile operators

and vendors to compare

– impact of EE features

– different architectures

› e.g. macro vs. HetNet vs. C-RAN deployment etc.


›

deployment etc.

› Strong political interest

– Industry need to show progress or face governmental mandates to reduce energy use

IntroductionChallenges to make a fair comparison› Comparing EE of networks between countries difficult:

– Population densities variation

– Geography – e.g. Netherlands vs Switzerland

– …

› Comparing EE of networks between standards difficult:– Different spectrum bands and bandwidths

…


–

– …

› Comparing EE of networks carrying different traffic mix difficult– chatty traffic more power consuming than constant bit rate traffic

(video, voice)

– …?

A B

Outline

› Introduction






› Conclusion

Traffic evolutionTo

tal (U

L+

DL

) m

on

thly

tra

ffic

Pe

tab

yte

/m

on

th Data is the most interesting traffic for EE measurement


› Data traffic increases ~ doubles every year

› Voice traffic increases or almost constant

To

tal (U

L+

DL

) m

on

thly

tra

ffic

Pe

tab

yte

/m

on

th

Source: [1] Ericsson; Akamai, Q4 2011 State of the Internet Webinar May 10, 2012http://evt.dispeak.com/akamai/soti/soti-may12-a/Q4-2011-SotI-Webinar.pdf

Data Traffic volume per Application and usage; iPhone

Other

Email

Social netw.

Web

��

Encrypted

iTunes

Tra

ffic

vo

lum

e s

ha

res

100%


Energy consumption depends on Traffic mix and Clients

��

iTunes

Web radio

HTTP video

Tra

ffic

vo

lum

e s

ha

res

0%

Heavy usersLight users

Source: [1] Ericsson; Akamai, Q4 2011 State of the Internet Webinar May 10, 2012http://evt.dispeak.com/akamai/soti/soti-may12-a/Q4-2011-SotI-Webinar.pdf

› PC dongles, Pads and Android phones show similar but

some different traffic mix (ref. [2], [3] and [4])

Energy consumption depends on Traffic mix and Clients

How measure EE?

› Energy / Bit › Energy / Bit / Hz

Control channel

overhead


Small � High EE

Large � Low EE

DataCtrl

DataCtrl

10 20 MHz BW

Network A

Network B

The EARTH utility function provides a method for comparing

How measure EE?

Performancerequirements

QoS requirements?


AddQoE requirements?

› Throughput only one important component for the end user QoE

› Applications have very different service quality demands– Voice & video have strict latency requirement

– Other applications are delay tolerant –pushed content to UE etc.

QoE KPIs are important indicators of performance

Outline alternative

› Introduction






› Conclusion

QoS measurement scenarioThroughput based KPI

Internet

Throughput KPI

Counters

EventsOSS/NMS


› Measured data from RNC, RBS, SGW collected by OSS

› Throughput data is selected

RANPacket

core

Internet

Server

Counters

Events

QoE measurement scenarios

QoE model

Internet

KPI

OSS/NMS

1,2 Measurement points

3

User

reporting

4 Standardization


RANPacket

core

Internet

Server

1 1 2

3

Deep

Packet

Inspection

Counters

Events

Terminal

reporting

Network probing

Internet

OSS/NMS

1

QoE model

KPI


› Deep Packet Inspection, DPI of IP packet header for

application identification

› Probes needs to be updated with new application

RANPacket

core

Internet

Server

Probing implementations› GGSN DPI

– IP header in GGSN is checked

– Filtered packets sent to server for analyses

› DPI on separate node

GiGn

1


GGSN

Packet filtering activated,

Juniper J-Flow software (example)

Server with

Traffic Analyzer software

GiGn

Examples:

Server

…

DPI on any

interface – one

vendor example

Terminal reporting

OS

App

Internet

2Special designed App

for EE measurements?QoE model

KPI

OSS/NMS


› 3GPP MDT, Minimization of Drive Test – QoS reporting part

– User consent needed, might imply low penetration

– Measurement parameters suitable for EE measurements?

RANPacket

core

Internet

Server

› App designed for EE measurements – standardized?

User reporting

OS

App

Internet

3Special designed App for

User reported QoE?QoE model

KPI

OSS/NMS


› 3GPP MDT, Minimization of Drive Test – QoS reporting part

– User consent needed, might imply low penetration

– Measurement parameters suitable for EE measurements?

RANPacket

core

Internet

Server

› App designed to support user reported QoE?

Compare with MOS measurements!

Video QoE example: ITU-T modelP.1201 (P.NAMS)4

Fast accessenables short time to fill buffer

Re-buffering?

MbpsVideo service


Time

P.1201 model used to estimate video QoE

Parameters in the model (subset)› Video Codec (H.264, etc.)

› Format: Video resolution QCIF, HVGA, …)

› Bitrate: coded video bitrate (kbps)

› Packet loss information: lost packets for video sample (~16 sec).

› Buffering events: Freezing of the video

Outline

› Introduction



› QoE measurement scenarios



› Conclusion

Conclusion

› Quality of Experience KPIs are possible to be used

› Several measurement alternatives are available

› Video QoE models are being standardized – other


› Video QoE models are being standardized – other

applications in process

› A KPI to be used for Energy Efficiency measurement of

networks needs to be agreed upon

QoE KPIs are important indicators of performance for Network Energy Efficiency measurements

References

› [1] Akamai, The State of the Internet 4th Quarter, 2011 reportVolume 4, Number 4, http://www.akamai.com/stateoftheinternet/

› [2] Akamai, Q4 2011 State of the Internet Webinar May 10, 2012http://evt.dispeak.com/akamai/soti/soti-may12-a/Q4-2011-SotI-Webinar.pdf

› [3] Ericsson: Ericsson predicts Mobile Data Traffic to grow 10-fold by 2016Press Releases, Reports Nov 7, 2011: New report from Ericsson says mobile data traffic will grow 10-fold between 2011 and 2016, mainly driven by video Mobile broadband.


grow 10-fold between 2011 and 2016, mainly driven by video Mobile broadband.

› [4] Ericsson: Traffic and Market Data Report: On the pulse of the

Networked Society, http://hugin.info/1061/R/1561267/483187.pdf

› [5] ITU-T Study Group 12 (Study Period 2009-2012) Performance, QoS

and QoE http://www.itu.int/ITU-T/studygroups/com12/index.asp

Thank You!


Thank You!


KPIs for Energy Efficiency Network measurements - ETSI · PDF fileKPIs for Energy Efficiency Network measurements ... vs. C-RAN deployment etc. Energy Efficiency KPIs for ... QoE KPIs

Documents