1
Right measures and a quick benchmark of “Hyper
scale” Indian players
Nagu Gopalakrishnan
Director & Principal consultant, Automation & QoE
Solutions
nagu@tataelxsi.com
Award winning UX and design team
studios in London, Bangalore, Pune
DesignTechnology
6000+ technologists, engineers
Global R&D centers and labs
Content maybeking,Customer is the
kingmaker.
S C A LE & D I V ER S I T Y C H A LLEN G ES
Extremelydiversecontent consumption
preferences and varying last mileconnectivity
R EV EN U E P ER S U BS C I BER
DOMOREWITHLESS. Customer basemay
behigh,But revenueper subscriber is not on
par with western world. 5Bn+USDprojected
by2023. Not to mention churn.
Extremelydiversecontent consumption
preferences and varying last mileconnectivity
R EV EN U E P ER S U BS C I BER
DOMOREWITHLESS. Customer basemay
behigh,But revenueper subscriber is not on
par with western world. 5Bn+USDprojected
by2023. Not to mention churn.
D EFI N I N G EX P ER I EN C E K P I S
QoEshould beleading and lagging.
ExperienceKPIs areto controlled and not
just monitored.
Content maybeking,Customer is the
kingmaker.
S C A LE & D I V ER S I T Y C H A LLEN G ES
Extremelydiversecontent consumption
preferences and varying last mileconnectivity
R EV EN U E P ER S U BS C I BER
DOMOREWITHLESS. Customer basemay
behigh,But revenueper subscriber is not on
par with western world. 5Bn+USDprojected
by2023. Not to mention churn.
SomeExperienceKPIs measured and benchmarked
15 0 + M I LLI O N C U S T O M ER S
Content maybeking,Customer is the
kingmaker.
S C A LE & D I V ER S I T Y C H A LLEN G ES
Extremelydiversecontent consumption
preferences and varying last mileconnectivity
Quick Look at Indian Market
Performance monitoring of OTT Apps
OTT performance monitoring covers two aspects
Performance of video
streams
Performance of Applications
or application streaming
State and status of the art today
Streaming
performance
App performance
Challenges
• Mostly agent based and hence often
a lagging indicator.
• Often proactive control as a CI/CD
gate does not exist
• Issues often blamed on to last mile
connectivity
• Environment simulation is often not
scalable
Challenges
• Often differing dev, test and deployment
environment
• Heavy on browser dependent testing and
light on environment dependencies
• Big screen experiences are often
overlooked
• Automated performance measure in a CI
environment is often not used
Scale of testing is often limited to very few real
devices
What are the right measures and how do we
measure?
“Experience” Issues Trends across OTT
markets
40%
Never
triaged
Not reproducible
category
35%
Blame Game
Blamed on
Network/wifi/CDN
issues
25%
Triaged
and in backlog
Trend focuses on OTT bugs from the field across 30 different video players.
Monitoring is always a lagging indicator
Often consumed by Marketing, Operations and Strategy teams
Enabling experiences across the spectrum
Factors contributing Experience
Streaming conditions Network conditions End device conditions
Encoder related
Codec Related
Streamer Related
Packager Related
Others….
WAN Latency
Connectivity Medium
CDN Related
Others..
Device Type & Settings
Device Configuration
Player implementation
App implementation
Battery %
OS Video handling
Others…
What can we quantify, which can be
controlled?
Life cycle of a IP video stream
Criticality of highly accurate performance measures – Measured using
FalconEye HPAPIs
Benchmarking three OTT players
50Mil+ downloads 100Mil+ downloads 50Mil+ downloads
AVOD, SVOD AVOD, SVOD AVOD, SVOD
0
2
4
6
8
10
12
400kbps 800kbps 3000kbps 4000kbps
TIME(SECONDS)
BANDWIDTH
Video Start up Time
HOTSTAR Zee5 SonyLiv
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
400kbps 800kbps 3000kbps 4000kbps
Blackframes
HOTSTAR Zee5 SonyLiv
0
0.005
0.01
0.015
0.02
0.025
400kbps 800kbps 3000kbps 4000kbps
Buffering Ratio
HOTSTAR Zee5 SonyLiv
Hotstar Zee5 SonyLiv
AVG BW=91827, RES=250x140 BW=200387, RES=180x144 AVG BW=206334, RES=256x144
AVG BW=144235, RES=320x180 BW=600822, RES=300x240 AVG BW=313261, RES=384x216
AVG BW=276158,RES=320x180, BW=799218, RES=450x360 AVG BW=463751, RES=512x288
AVG BW=370963,RES=416x234, BW=1197145, RES=600x480 AVG BW=651625, RES=640x360
AVG BW=551837, RES=640x360 BW=2485460, RES=900x720 AVG BW=967215, RES=960x540
AVG BW=777657
RES=720x404
AVG BW=1287316,
RES=1280x720
AVG BW=1358156,
RES=1280x720
AVG BW=2072762
RES=1280x720
AVG BW=2758412
RES=1920x1080,
ABR Functionality
So what does those graphs say?
More focused on lower
bandwidths
Poor player performance
Needs optimisation for
higher bandwidths
High spend on CDN and
storage costs
Higher the bandwidth better the
performance
Needs finetuning for
streaming performance
Focus on content quality
Inconsistent player
performance
Poor streaming
performance
14
• App Launch Time
• Guide / Metadata
profiling
• Network Connect
profiling
• Navigation profiling
• Transitions
profiling
• Session Error Rates
• Minimum Time for
Stable Usage
Network QoE
Video QoE
Player QoE
App QoE
• Video Start Time
• Video Start Failure
Rate / Frequency
• Video Buffering Ratio
/ Frequency / Count /
Duration
• Video Playback
Failure Rate /
Frequency
• Channel Change
Time
• Video Restart Time
• Video Freeze
• Audio Freeze
• Macroblocking
• Pixelation
• Blurriness
• Video Blackouts
• Single Color
Freeze
• Video MOS
• Video Bitrate
• Audio Bitrate
• Manifest Request
• Download Rate
• Bandwidth Throttle
Performance
Measure
• Buffer Fill Measure
• Player ABR efficiency
KPIs for lifecycle of IP Video apps
Not just monitor, but control experiences
nagu@tataelxsi.com
Enabling OTT services to control content
consumption experiences

HyperScale India OTT Operators and Right KPIs

  • 1.
    1 Right measures anda quick benchmark of “Hyper scale” Indian players Nagu Gopalakrishnan Director & Principal consultant, Automation & QoE Solutions nagu@tataelxsi.com
  • 2.
    Award winning UXand design team studios in London, Bangalore, Pune DesignTechnology 6000+ technologists, engineers Global R&D centers and labs
  • 3.
    Content maybeking,Customer isthe kingmaker. S C A LE & D I V ER S I T Y C H A LLEN G ES Extremelydiversecontent consumption preferences and varying last mileconnectivity R EV EN U E P ER S U BS C I BER DOMOREWITHLESS. Customer basemay behigh,But revenueper subscriber is not on par with western world. 5Bn+USDprojected by2023. Not to mention churn. Extremelydiversecontent consumption preferences and varying last mileconnectivity R EV EN U E P ER S U BS C I BER DOMOREWITHLESS. Customer basemay behigh,But revenueper subscriber is not on par with western world. 5Bn+USDprojected by2023. Not to mention churn. D EFI N I N G EX P ER I EN C E K P I S QoEshould beleading and lagging. ExperienceKPIs areto controlled and not just monitored. Content maybeking,Customer is the kingmaker. S C A LE & D I V ER S I T Y C H A LLEN G ES Extremelydiversecontent consumption preferences and varying last mileconnectivity R EV EN U E P ER S U BS C I BER DOMOREWITHLESS. Customer basemay behigh,But revenueper subscriber is not on par with western world. 5Bn+USDprojected by2023. Not to mention churn. SomeExperienceKPIs measured and benchmarked 15 0 + M I LLI O N C U S T O M ER S Content maybeking,Customer is the kingmaker. S C A LE & D I V ER S I T Y C H A LLEN G ES Extremelydiversecontent consumption preferences and varying last mileconnectivity Quick Look at Indian Market
  • 4.
    Performance monitoring ofOTT Apps OTT performance monitoring covers two aspects Performance of video streams Performance of Applications or application streaming
  • 5.
    State and statusof the art today Streaming performance App performance Challenges • Mostly agent based and hence often a lagging indicator. • Often proactive control as a CI/CD gate does not exist • Issues often blamed on to last mile connectivity • Environment simulation is often not scalable Challenges • Often differing dev, test and deployment environment • Heavy on browser dependent testing and light on environment dependencies • Big screen experiences are often overlooked • Automated performance measure in a CI environment is often not used Scale of testing is often limited to very few real devices
  • 6.
    What are theright measures and how do we measure?
  • 7.
    “Experience” Issues Trendsacross OTT markets 40% Never triaged Not reproducible category 35% Blame Game Blamed on Network/wifi/CDN issues 25% Triaged and in backlog Trend focuses on OTT bugs from the field across 30 different video players. Monitoring is always a lagging indicator Often consumed by Marketing, Operations and Strategy teams
  • 8.
  • 9.
    Factors contributing Experience Streamingconditions Network conditions End device conditions Encoder related Codec Related Streamer Related Packager Related Others…. WAN Latency Connectivity Medium CDN Related Others.. Device Type & Settings Device Configuration Player implementation App implementation Battery % OS Video handling Others… What can we quantify, which can be controlled?
  • 10.
    Life cycle ofa IP video stream Criticality of highly accurate performance measures – Measured using FalconEye HPAPIs
  • 11.
    Benchmarking three OTTplayers 50Mil+ downloads 100Mil+ downloads 50Mil+ downloads AVOD, SVOD AVOD, SVOD AVOD, SVOD
  • 12.
    0 2 4 6 8 10 12 400kbps 800kbps 3000kbps4000kbps TIME(SECONDS) BANDWIDTH Video Start up Time HOTSTAR Zee5 SonyLiv 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 400kbps 800kbps 3000kbps 4000kbps Blackframes HOTSTAR Zee5 SonyLiv 0 0.005 0.01 0.015 0.02 0.025 400kbps 800kbps 3000kbps 4000kbps Buffering Ratio HOTSTAR Zee5 SonyLiv Hotstar Zee5 SonyLiv AVG BW=91827, RES=250x140 BW=200387, RES=180x144 AVG BW=206334, RES=256x144 AVG BW=144235, RES=320x180 BW=600822, RES=300x240 AVG BW=313261, RES=384x216 AVG BW=276158,RES=320x180, BW=799218, RES=450x360 AVG BW=463751, RES=512x288 AVG BW=370963,RES=416x234, BW=1197145, RES=600x480 AVG BW=651625, RES=640x360 AVG BW=551837, RES=640x360 BW=2485460, RES=900x720 AVG BW=967215, RES=960x540 AVG BW=777657 RES=720x404 AVG BW=1287316, RES=1280x720 AVG BW=1358156, RES=1280x720 AVG BW=2072762 RES=1280x720 AVG BW=2758412 RES=1920x1080, ABR Functionality
  • 13.
    So what doesthose graphs say? More focused on lower bandwidths Poor player performance Needs optimisation for higher bandwidths High spend on CDN and storage costs Higher the bandwidth better the performance Needs finetuning for streaming performance Focus on content quality Inconsistent player performance Poor streaming performance
  • 14.
    14 • App LaunchTime • Guide / Metadata profiling • Network Connect profiling • Navigation profiling • Transitions profiling • Session Error Rates • Minimum Time for Stable Usage Network QoE Video QoE Player QoE App QoE • Video Start Time • Video Start Failure Rate / Frequency • Video Buffering Ratio / Frequency / Count / Duration • Video Playback Failure Rate / Frequency • Channel Change Time • Video Restart Time • Video Freeze • Audio Freeze • Macroblocking • Pixelation • Blurriness • Video Blackouts • Single Color Freeze • Video MOS • Video Bitrate • Audio Bitrate • Manifest Request • Download Rate • Bandwidth Throttle Performance Measure • Buffer Fill Measure • Player ABR efficiency KPIs for lifecycle of IP Video apps
  • 15.
    Not just monitor,but control experiences
  • 16.
    nagu@tataelxsi.com Enabling OTT servicesto control content consumption experiences