meeting the ott challenge
TRANSCRIPT
Meeting the OTT challenge
Martin Geddes
Martin Geddes Consulting Ltd
© 2013 All Rights Reserved
What has to change?
NOW FUTURE
PURPOSE-FOR-FITNESS
FITNESS-FOR-PURPOSE
Core thesis
1. Networks are (option) trading spaces – That match supply and demand across all timescales
2. Your business is statistical multiplexing for fun and profit – Supply and demand meet here, and trades are made
3. Success primarily depends on how well you do this – Regardless of the (OTT) business model on top or who pays
4. Your current business is mathematically unsustainable – Because you have not taken full control over your network trading space
5. There is a way to take control – Get away from supply-push “bandwidth” approach & purpose-for-fitness
6. Move to a sustainable demand-driven “quality” model
What to do?
1. Characterise demand and create fit-for-purpose supply
2. Align your design, marketing, operations to deliver
3. Execute to create differentiation in cost and QoE
4. Enable new OTT business models
The Facts
Situation
OTT voice and messaging are hurting telephony and SMS revenue
Selling data speed and mechanisms
Value is measured in data volume
Revenue model: Proportional to average volumetric demand
Cost model
Size to peak demand
Planned upgrades Volume-driven capacity
planning rules
Unplanned upgrades Driven by churn and complaints
Key properties of data demand
• User have a sense of entitlement
– Want properties of circuits
– Uncontended, on-demand un-impaired capacity
• Ability to attach any device or application
– Demand shocks can and do happen (iPhone, Olympics, emergency events, etc.)
• Distribution of use is shifting
– Not just the average; peaks are getting “peakier”
Key properties of data supply offer
• One-size-fits-all: Single class of service
• One-sided market: End user pays
– No “toll free” data or upstream revenue
• No quality assurance or performance SLAs
• Little visibility of actual user experience
Supply-push model: Purpose-for-fitness
The market is evolving
• Rapid growth in demand
– SaaS/cloud, mobile workers, tablets, automotive, small cells, M2M, smart grids, etc.
• These require new supply capabilities
– Very different cost and quality profiles
The market is evolving
• Government and regulatory focus shifting to “digital dividend”
– Tackling economic/social issues
– It’s not going to be about negotiating roaming and termination rates in future
All operators are facing tough questions
1. How to sustain voice and messaging revenue and differentiation positioning?
2. How to relate to OTTs (block, bundle, ignore, service, join in, partner…)?
3. How to address growing market needs at an affordable cost?
What’s wrong
Complication
Speed (and volume) are not value
Dangerous myth: More Speed is Always Better
Contention exists!
Need to consider variability, not just speed.
Source: Predictable Network Solutions Ltd
Black cygnets: small “bad coincidences” create bad experiences
These coalesce under high load
And create ever more ‘black swan’ application failures
The application
Hierarchy of Need
3. Reasonable bounds on loss and delay
2. Sufficient stationarity
1. Sufficient capacity
Note: exact requirements are application-dependent
So 4G won’t solve your problems
Downstream delay over a 3G connection – 4G doesn’t change this unwanted variability
Too much variability for TCP to work well.
Source: Predictable Network Solutions Ltd
What you need to know
Some theory
Capacity demand
TWO sources of network demand
Schedulability demand
Capacity demand LOW HIGH
Feasible Infeasible
MAX CAPACITY
TWO fundamental resource limits
Feasible
MAX SCHEDULABILITY Sc
he
du
lab
ility
d
em
and
Infeasible
LOW
HIGH
Problem
Schedulability demand is growing fast
VoIP, gaming, 2-way video, UC, HTML5 web, WebRTC…
Problem
Solving schedulability issues (i.e. non-stationarity)
with capacity is inefficient and ineffective
Problem
Monoservice network means costs track the
worst-case schedulability limit of loading
Summary so far
• “Bandwidth” is your current input and output – This is not a good proxy for fitness-for-purpose
– Other factors also matter to QoE
• Revenue is from fit-for-purpose experiences – But you have stopped paying attention to user needs
– Dependability is not on sale, at any price
• Costs are being driven by schedulability issues – Every flow has the same cost structure as your most
quality-demanding users/flows
– But schedulability isn’t part of costing & ops model
The consequence
Undesirable future
Telecoms is a capital killer ($60bn/year shortfall, every year)
Source: PwC http://www.pwc.com/en_GX/gx/communications/publications/assets/pwc_capex_final_21may12.pdf
Failure of technology to keep
up with ever rising demand
forces shorter upgrade cycles
Rising load makes
service quality fall,
forcing upgrades
Serv
ice
Qu
alit
y
Time
Un
dep
reci
ated
Ass
et V
alu
e
Time
Mathematically unsustainable
More, more, more (aka 2G/3G/4G/5G cycle of doom)
More supply
More elastic demand
Faster saturation of backhaul
More non-stationarity
More complaints and churn
Race to the bottom?
The alternative
Desirable future
What do we want?
• Demand – increased benefits
– Able to match a wide range of quantity, quality and cost needs
– Can package offers to fit segments
• Supply – decreased costs
– Costs scale sub-linearly with users
– Predictable in-life operational costs
Packaged (OTT) cloud applications
• Available when and where you need it
• Right quantity and quality
• At a cost you can afford
• Easy to consume
How to get there?
The Question
The big question
How can we exploit the trades (and demand-shift by scheduling)
and match supply to demand to create the
right QoE and cost trade-offs?
Then, given that capability, what should our OTT strategy be?
What do I need to do?
The Answer
Bandwidth Quality Need to frame the problem
differently to make it soluble
What has to change?
NOW FUTURE
PURPOSE-FOR-FITNESS
FITNESS-FOR-PURPOSE
Focus on enabling outcomes – not shifting data Make bad experiences rare(r).
Lower cost of delivering good experiences.
TELCO END USER
Manage benefits, costs and risks across supply chain
BENEFIT
COST
RISK (failed call)
Made the sales call
Price of phone call
Didn’t make sale
Second car
Revenue
Tin, opex
SLA breach or churn
Unplanned capacity upgrade
Time wasted
Reputational loss
INSURANCE Contingency fund (lawsuit, PR)
Frustration
Excess risk has to be (self-)insured
Manage QoE risk through network resource “trades”
The “tails” of loss and delay + their structure
are what cause application QoE failure, and whose mitigation
drives cost.
Source: Predictable Network Solutions Ltd
Lower cost of good experiences by time-shifting delay-insensitive traffic
• Reduce cost by lowering peaks
– Currently encouraging people not to time-shift.
– Users behave in a predatory way.
• Mark bulk traffic
– Cheaper to post bulk mail if pre-sorted.
Microseconds to minutes
Peak demand
Summary: Do’s and Don’ts
• Do: – Explore the nature of the market – who is paying for what?
– Think systemically; optimise globally
– Become aware your implicit bandwidth thinking and its dangers
– Exploit packet-based statistical multiplexing
• Don’ts: – Focus on supply inputs and volume; it’s about outcomes
– Mistake trades for QoS
– Sell circuits – you will be arbitraged (cf ISPs in 1990s)
– Think you can solve this without differentiation
It’s all about the trading space
The logistics companies out-competed the shipping companies because they controlled the resource
trading space
Get in touch to discuss the necessary changes to network design,
operations, marketing & product management to meet OTT challenge
Martin Geddes [email protected]