Bandwidth demand forecasting - IEEE 802ieee802.org/3/ad_hoc/ngepon/public/sep14/harstead_ngepon_01a_091… · Ed Harstead & Randy Sharpe, members of Fixed Networks CTO, Alcatel-Lucent

COPYRIGHT © 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED.

1 1

Ed Harstead & Randy Sharpe, members of Fixed Networks CTO, Alcatel-Lucent Ottawa IEEE 802.3 NGEPON Interim meeting 8 September 2014

Bandwidth demand forecasting (for TR section 4.2.2)


2 2

Residential aggregate bandwidth demand forecast

100

1000

10000

100000

2014 2016 2018 2020 2022 2024

Agg

rega

ted

dem

and

(Mb/

s)

99%-ile aggregated demand: "heavy" scenario, 100% unicast video

512 subscribers

256 subscribers

128 subscribers

64 subscribers

32 subscribers

16 subscribers

8 subscribers

For PON network dimensioning, add bandwidth headroom above these numbers, equivalent to the maximum service level offered Why: subscribers will need to pass a speed test at peak hour traffic


3 3

Aggregate bandwidth demand model

100

1000

10000

100000

2014 2016 2018 2020 2022 2024

Agg

rega

ted

dem

and

(Mb/

s)

99%-ile aggregated demand: "heavy" scenario, 100% unicast video

512 subscribers

256 subscribers

128 subscribers

64 subscribers

32 subscribers

16 subscribers

8 subscribers

Peak hour sustained and burst bandwidth demands are modeled independently.

1. Sustained bandwidth demand component: • Video traffic is modeling using Monte Carlo

techniques. Other sustained demand (e.g. voice) is neglected

• All of managed linear pay-TV, managed VoD and OTT Internet video (e.g. Netflix, YouTube) are included

• No multicasting all video (managed and OTT) is unicasted in-band: worst case scenario

2. Burst bandwidth demand component • per subscriber averages, plus

• maximum burst demand that a single subscriber can generate

•More detail on inputs in the backup slides •Manuscript has recently been submitted to IEEE Communications Magazine


4 4

100

1000

10000

2014 2016 2018 2020 2022 2024

Band

wid

th (M

b/s)

Maximum service level offering

WDM PON (10 Gb/s/λ)

10G EPON (32 subscribers)

GPON (32 subscribers)

WDM PON (1 Gb/s/λ)

Maximum service level that can be offered

Equivalent to headroom left over after peak hour traffic is subtracted out:

Notes on WDM PON: – WDM PON @1 Gb/s/λ loses to GPON over next 10 years. – WDM PON must support 10 Gb/s/λ just to beat (barely) today’s 10G EPON

Based on methodology in: E. Harstead, R. Sharpe, “Future Fiber-To-The-Home bandwidth demands favor Time Division Multiplexing Passive Optical Networks”, IEEE Communications Mag., Nov. 2012.


5 5

Input for section 4.3: Bit Rate Trends Question: Is this section about bandwidth offered or bandwidth demanded?

Nielsen's Law of Internet Bandwidth states that offered premium connection speeds to U.S. users has grown 50% YoY.

From Jakob Nielsen http://www.nngroup.com/articles/law-of-bandwidth/

Nielsen’s Law does not say anything about: – bandwidth demand – offered speeds on fiber networks (off by one order of magnitude)

Nielsen’s Law ignores the bandwidth carrying managed TV service (e.g. video QAMs)

1 Gb/s

GPON 2011 (EPB, ZON, HKBB)


6 6

Backup


7 7

Video streams transported: “heavy scenario”

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0 1 2 3 4 5 6 7 8

Streams/sub probability

Probability distribution of number of streams per subscriber

Average (λ) =1.8

0.000.100.200.300.400.500.600.700.800.901.00

Video resolution probability

SD

720p60 HD

1080p60 HD

2160p60 ("4k")

4320p120 ("8k")

Stream resolution probability distribution forecast

3D content: 7.5% of HD streams


8 8

0.000.100.200.300.400.500.600.700.800.901.00

Video resolution probability

SD

720p60 HD

1080p60 HD

2160p60 ("4k")

4320p120 ("8k")0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0 1 2 3 4 5 6 7 8

Streams/sub probability

Video streams transported: “moderate scenario”

Probability distribution of number of streams per subscriber

Average (λ) =1.2 Stream resolution probability distribution forecast

3D content: 5% of HD streams


9 9

Residential aggregate bandwidth demand forecast “Moderate” scenario

10

100

1000

10000

2014 2016 2018 2020 2022 2024

Agg

rega

ted

dem

and

(Mb/

s)

99%-ile aggregated demand:"moderate" scenario, 100% unicast video

512 subscribers

256 subscribers

128 subscribers

64 subscribers

32 subscribers

16 subscribers

8 subscribers

No headroom included


10 10

Encoded video bit rate forecasts

0

10

20

30

40

50

60

2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024

Enco

ded

bit r

ate

(Mbp

s)

4320p120 ("8k")

2160p60 ("4k")

1080p60 HD

720p60 HD

SD

• Chart forecasts nominal bit rates for good quality video • HD and UHD bit rates are multiplied by

• 1.275 quality factor for higher quality (heavy scenario) • 0.85 quality factor for higher bandwidth efficiency (moderate scenario)

• 3D content: 1.4x higher bit rate than 2D


11 11

Burst traffic demand component

Cox reported an average of 400 kb/s of downstream DOCSIS traffic per subscriber during peak hour in 2014[1]. This is consistent with private data from other operators.

– Assume that to first order it is all Internet traffic (neglect managed VoD) – Based on Sandvine measurements[2] 36% of that traffic is bursty: 144

kb/s average peak hour bursty traffic per subscriber on a DOCSIS network.

– Traffic could be higher on fiber networks: multiply by 1.5x: 216 kb/s per subscriber

– We grow this number by 21% YoY based on inputs from the Cisco VNI[3] and Pyramid[4] and multiplying by 1.5x

[1] J. Finkelstein, A. Bernstein, S. Parikh, “SDN and NFV for Cable Networks”, 2014 Spring Technical Forum (NCTA, SCTE and CableLabs), Los Angeles, CA. [2] Sandvine, ”Global Internet Phenomena Report: 1H 2014”, https://www.sandvine.com/trends/global-internet-phenomena/ [3] Cisco, “Cisco Visual Networking Index: Forecast and Methodology, 2012–2017”, http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-481360.pdf, May 2013. [4] Pyramid Research, “Western Europe Fixed Forecast Pack 2Q2013” and “North America Fixed Forecast Pack, 2Q2013”, June 2013.

https://www.sandvine.com/trends/global-internet-phenomena/

http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-481360.pdf


12 12

Streaming video bandwidth demand, single subscriber: upper bound

0

50

100

150

200

250

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025

Ban

dwid

th d

eman

d (M

b/s)

15% YoY upper bound

4

1080p60

720p60

SD

2160p60

4320p120

Assumes no significant impact of 3D video

Based on methodology in: E. Harstead, R. Sharpe, “Future Fiber-To-The-Home bandwidth demands favor Time Division Multiplexing Passive Optical Networks”, IEEE Communications Mag., Nov. 2012.


13 13

0

100

200

300

400

500

600

700

800

900

1000

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

20000

2012 2013 2014 2015 2016 2017

Traffic per subscriber (kb/s)

Tota

l Tra

ffic

(Cis

co V

NI)

(PB/

mon

th)

North America Consumer Internet Traffic

Total traf f ic Traf f ic per subscriber

Bandwidth demand: Internet traffic forecasts?

Internet traffic forecasts miss IPTV, so underestimate access network demand

Internet video is the driver for growth But subscriber traffic is still measured in 100s of kb/s!

Traffic forecast data from Cisco Visual Networking Index, 2013

Subscriber forecast data from Pyramid Research. Fixed Forecast Pack, 2013


14 14

Documents

Bandwidth demand forecasting - IEEE 802ieee802.org/3/ad_hoc/ngepon/public/sep14/harstead_ngepon_01a_091… · Ed Harstead & Randy Sharpe, members of Fixed Networks CTO, Alcatel-Lucent