Optimising Mobile Content Mobile Backhaul Asia 2012, Bangkok 28.3.2012 Martin Prosek, VAS Platforms Development Manager Telefónica Czech Republic
Telefónica – Global Operator
1
About Telefónica Czech Republic
Fixed and mobile voice and data, IPTV Operated under commercial brand O2
2
Overview
01 Mobile data traffic and networks – current state 02 Data traffic optimisation: what and why? 03 Impact on customer experience 04 Optimisation costs 05 Summary and recommendations
Disclaimer: The opinions of the author expressed in this document do not necessarily state or reflect those of Telefónica company 3
Mobile Data Traffic Surge
Mobile data use skyrockets everywhere Telefónica CZ is not an exception Technology availability is not enough
Cheap data packages launch HSDPA launch
The pricing drives the usage
4
Reasons for Traffic Increase
New devices – smartphones, tablets, connected PCs… New mobile access technologies – HSDPA, HSPA+, LTE… New services – OTT…
5
Impact of Increased Data Traffic
Network overload – data congestion Connection unreliability or unavailability
150 %
100 %
Average
Monday
Tuesday
6
Impact on customer
Structure of Mobile Network
Mobile Backhaul
Mobile Backhaul
Many places have bottlenecks Devices have big influence
DPI
DHCP
7
AAA
Optimization
Impact on Mobile Backhaul Overhead issues
3G 3G 2G 2G
. . .
Optimization
Internet
Small increase Link overload
2G Great savings potential when leased infrastructure is used
2G
RAN
Backhaul
8Core
NW
IP NW
Network Overload Prevention
Right pricing Network capacity upgrade (more cells – e.g. femto, new
technology…) Traffic shaping (policy control – FUP, differentiated services…) Traffic offloading (WiFi…) Traffic data optimisation
What traffic?
How optimise?
9
Downlink Data Traffic Structure
HTTP
YouTube
Flash video
MPEG video
10
RTMP
Uplink – Downlink Scale Comparison
Downlink
Uplink
11
Uplink Data Traffic Structure
HTTP
BitTorrent
Skype
SSL
12
Facebook
Ways to Optimise Data Traffic
Traffic optimisation •
On protocol level
Content optimisation • •
Data squeezing – compression, re-encoding… Caching – CDN…
13
Traffic Optimization
Protocol optimization (every vendor has own name…) • • • •
VTP TCP+ MDI (Macara Dynamic Interleaving) …
Basically it is improvement of current transport protocols
(TCP retransmissions, packets defragmentation, headers compression…)
Minimization of transport overhead and latency 14
Content Optimisation
Why it might work? • • •
Fixed Internet and Mobile Internet are overlapping Content providers are usually not distinguishing between these channels and keep the content in form suitable for fixed Internet Thus the transferred data are not optimized for mobile access
Content optimization should • •
Bring better customer experience by quicker access to desired content Prevent network congestion
Only some content might be optimized 15
Content Optimisation Options
Data squeezing (possibly without decreasing quality) • • •
Removal of redundant data Compression Transcoding (codecs, multibitrate…)
Data transfer •
Issues with content owner rights?
JIT (Just-In-Time) or buffer tuning
16
Optimisation of HTTP Browsing
Protocol optimisation •
HTTP 1.1 keep alive, no unnecessary refreshes (304 – caused by Etags and Last-modified headers), use browser caching
Transfer compression •
Gzip encoding saves about 20 % of data (HTML, JS, CSS…)
Avoid scattering (too many small files) •
Use multipart or images stripping
Downsized images • • •
Efficient compression (GIF, PNG, JPEG…) Lossy squeezing (resolution, number of colours, encoding parameters…) Avoid scaling by HTML parameters, invisible parts should be cropped 17
Optimisation of Video
Video •
Effective codecs
› Transcoding to lower quality, lower framerate, less keyframes…
• • •
Different bandwidths (preferrably multibitrate or smooth streaming) Just-in-time delivery …
18
Optimisation without Impact on User Experience
Loseless compression •
Impossible to recognize
Lossy compression •
Lower quality is usually invisible on small screen devices
19
Quality Degradation
20
Quality Degradation
Visibility of quality degradation depends on user perception
Small screen hides problems 21
Where Network Based Content Optimisation Fails
VPN HTTPS P2P VoIP … Also where the content owner bans any modification! For signalling traffic 22
Caching, Optimisation Policy Control
Caching on the edge to Internet •
Almost standard thing, base of many CDNs
Depends on the type of content consumed by users!
Caching in RAN •
Quite risky, still not fully evolved (experimental…)
Optimisation can be enforced on congested sites •
Policy control can be used for this
23
Issue with congested cell Identification!
Where is the Best Place for Optimisation?
Mobile Operator End-user Device
Optimisation Platform
Content Provider
Big cost 24
Costs of Content Optimisation
Who should pay the cost of content optimisation solution at operator?
Customers? •
They will not pay premium cost for it.
Content providers? • •
They can optimise it already at their servers with lower cost. Possibility to use operators CDN with Content Optimisation.
Operator? •
If it would save any investment into network capacity…
25
Data traffic
Costs of the Content Optimisation
Total traffic without compression
Saved traffic
Compressed traffic “– Inves ” tmen t
Compressible traffic Uncompressible traffic “+” S a Postponed investment vin gs
Capa “ – ” city L icens e
Years
Savings > Investment + Capacity license ? 26
Operators’ Rationale for Deploying
Savings – not Customer experience – yes Commercial Potential •
When network is not just dumb pipe its „smart“ capabilities can be used to better collaborate with OTT providers
27
Summary
Real cost of congestion is in customer dissatisfaction Data traffic optimisation is viable method to fight
with
network overload
Mobile backhaul is heavily impacted by local congestion •
Optimisation is not a “silver bullet” but a good helper at least
Content
Optimisation competitors
can
28
help
differentiate
from
Thank you.
Devices – Which are the Source of Overload?
Traditional devices (feature phones) do not usually cause
overload PC (laptop) users with modems (USB dongles) demand the same services as over fixed network Smartphones are bigger problem for signalling than overload, especially due to applications that are not respecting mobile network specifics Smartphone/PC crossover = tablet – combination of drawbacks?
30
Devices Behaviour Smartphones are usually connected „always on“ but actively used for
short periods. They might cause signalling overload due to „push services“ but amounts of data are limited.
PCs (laptops) with modems or USB dongles/datacards have infrequent,
but high-volume consuming sessions with main focus on download. Tablet-style devices are similar to PCs but might be preferably used for bulk content downloads or video streaming, mostly in indoor locations. Gaming devices need both high speed and also low latency.
M2M
or telemetry devices have varying usage dynamics – smart meters and health-monitoring devices may have very low volumes of data traffic but need absolute priority and guarantees. Connected CCTV cameras and sensors have high uploads. 31
Data Traffic by Mobile Device
Every type of mobile device has different impact
Laptops are just x 200 in O2 CZ network iPhones are just x 10 in O2 CZ network 32
Devices Split in network in TEF CZ
Large amount of feature phones – especially Nokia 6300 Modems 4% Tablets 0% Smartphones 20%
Feature phones 76%
33
Devices Connecting on-line in TEF CZ
Feature phones are also connecting Modems 19%
Feature phones 24%
Tablets 1%
Smartphones 56% 34
Transferred Data Volume in TEF CZ
Modems are eating majority of the bandwidth Feature phones 1% Smartphones 15% Tablets 1%
Modems 83%
35
Signalling Load in TEF CZ
Smartphones are taking more than others Applications are connecting – application updates, messaging, advertising… Modems 15%
Feature phones 6%
Tablets 2%
Smartphones 77%
36
data download,
Data Traffic by Device in TEF CZ
This data is provided by
37
Devices Dynamics in TEF CZ
Market share reports do not give proper info about actual state in networks Operators measure it more accurate – but not publish Example from TEF CZ network
Google Android Apple iOS
200802 200806 200810 200902 200906 200910 201002 201006 201010 201102 201106 38
