Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript or download a copy.
Here to
describe new models of Internet Protocol (IP) video and time-managed IT financing
is Paul Larbey, Head
of the Video Business Unit at Nokia,
based in Cambridge, UK. The discussion is moderated by Dana Gardner, Principal Analyst at Interarbor Solutions.
Here are some excerpts:
Gardner: It seems that the video-delivery business is in upheaval. How are video delivery trends coming together to make it necessary for rethinking architectures? How are pricing models and business models changing, too?
It’s been
10 years since the creation of the Apple iPhone.
The iPhone sparked whole new device-types, moving eventually into the iPad. Not
only that, Apple underneath developed a lot of technology in terms of how you
stream video, how you protect video over IP, and the technology underneath that,
which we still use today. Not only did they create a new device-type and avenue
for us to watch video, they also created new underlying protocols.
Gardner: It sounds like a few things need to change between the original content’s location and those mobile screens and those customized user scenarios you just described. What underlying architecture needs to change in order to get us to 2027 safely?
It can be a private CDN product, which we have deployed for the last seven years, and we have a cloud digital video recorder (DVR) product, which is all about moving the storage capacity into the network. We also have a systems integration part, which brings a lot of technology together and allows operators to combine vendors and partners from the ecosystem into a complete end-to-end solution.
In the private CDN and the video applications, our goal is to use 100 percent of the storage all of the time to maximize the cache hit-rate. With the traditional IT payment model for storage, my application fundamentally breaks that. So having a partner like HPE that was flexible and could understand the application is really important.
Larbey: Yes, absolutely. That is becoming more and more important. If you go back to the very first the Internet video, you watched of a cat falling off a chair on YouTube. It didn’t matter if it was buffering, that wasn't relevant. Now, our tolerance just doesn’t exist anymore for buffering and we demand and expect the highest-quality video.
So there is a lag between the live event and what you are seeing on your IP stream, which is 30 to 40 seconds. If you are in an apartment block, watching a live sporting event, and your neighbor sees it 30 to 40 seconds before you, that creates a big issue. A lot of the innovations we’re now doing with streaming technologies are to deliver that same broadcast experience.
We do a lot with Sky in
the UK, which was also looking to migrate away from an Internet-based CDN service
into something in-house so they could take more control over it and improve the
users’ quality of experience.
Here are some excerpts:
Gardner: It seems that the video-delivery business is in upheaval. How are video delivery trends coming together to make it necessary for rethinking architectures? How are pricing models and business models changing, too?
Larbey: We sit here in 2017, but let’s look
back 10 years to 2007. There were a couple key events in 2007 that dramatically
shaped how we all consume video today and how, as a company, we use technology to
go to market.
 |
| Larbey |
It was also
10 years ago that Netflix began to first
offer a video streaming service. So if you look back, I see one year in which how
we all consume our video today was dramatically changed by a couple of events.
If we fast-forward,
and look to where that goes to in the future, there are two trends we see today
that will create challenges tomorrow. Video has become truly mobile. When we
talk about mobile video, we mean watching some films on our iPad or on our
iPhone -- so not on a big TV screen, that is what most people mean by mobile
video today.
The future is personalized
When you can take your video with you, you want to
take all your content with you. You can’t do that today. That has to happen in
the future. When you are on an airplane, you can’t take your content with you.
You need connectivity to extend so that you can take your content with you no
matter where you are.
Take the
simple example of a driverless car. Now, you are driving along and you are
watching the satellite-navigation feed, watching the traffic, and keeping the
kids quiet in the back. When driverless cars come, what you are going to be
doing? You are still going to be keeping the kids quiet, but there is a void, a
space that needs to be filled with activity, and clearly extending the content
into the car is the natural next step.
And the
final challenge is around personalization. TV will become a lot more
personalized. Today we all get the same user experience. If we are all on the
same service provider, it looks the same -- it’s the same color, it’s the same
grid. There is no reason why that should all be the same. There is no reason
why my kids shouldn’t have a different user interface.
There is no reason why I should have 10 pages of channels that I have to through to find something that I want to watch.
The user interface presented to me in the morning may be
different than the user interface presented
to me in the evening. There is no reason why I should have 10 pages of channels
that I have to go through to find something that I want to watch. Why aren’t all
those channels specifically curated for me? That’s what we mean by personalization.
So if you put those all together and extrapolate those 10 years into the
future, then 2027 will be a very different place for video.Gardner: It sounds like a few things need to change between the original content’s location and those mobile screens and those customized user scenarios you just described. What underlying architecture needs to change in order to get us to 2027 safely?
Larbey: It’s a journey; this is not a step-change.
This is something that’s going to happen gradually.
But if you
step back and look at the fundamental changes -- all video will be streamed. Today,
the majority of what we view is via broadcasting, from cable TV, or from a
satellite. It’s a signal that’s going to everybody at the same time.
If you
think about the mobile video concept, if you think about personalization, that
is not going be the case. Today we watch a portion of our video streamed over
IP. In the future, it will all be streamed over IP.
And that
clearly creates challenges for operators in terms of how to architect the
network, how to optimize the delivery, and how to recreate that broadcast
experience using streaming video. This is where a lot of our innovation is
focused today.
Gardner: You also mentioned in the case of
an airplane, where it's not just streaming but also bringing a video object
down to the device. What will be different in terms of the boundary between the
stream and a download?
IT’s all about intelligence
Larbey: It’s all about intelligence. Firstly,
connectivity has to extend and become really ubiquitous via technology such as
5G. The increase in fiber technology will dramatically enable truly ubiquitous
connectivity, which we don’t really have today. That will resolve some of the
problems, but not all.
But, by the
fact that television will be personalized, the network will know what’s in my
schedule. If I have an upcoming flight, machine learning can automatically
predict what I’m going to do and make sure it suggests the right content in
context. It may download the content because it knows I am going to be sitting
in a flight for the next 12 hours.
Gardner: We are putting intelligence into
the network to be beneficial to the user experience. But it sounds like it’s
also going to give you the opportunity to be more efficient, with just-in-time
utilization -- minimal viable streaming, if you will.
How does
the network becoming more intelligent also benefit the carriers, the deliverers
of the content, and even the content creators and owners? There must be an
increased benefit for them on utility as well as in the user experience?
Larbey: Absolutely. We think everything
moves into the network, and the intelligence becomes the network. So what does
that do immediately? That means the operators don’t have to buy set-top boxes.
They are expensive. They are very costly to maintain. They stay in the network
a long time. They can have a much lighter client capability, which basically
just renders the user interface.
The first
obvious example of all this, that we are heavily focused on, is the storage. So
taking the hard drive out of the set-top box and putting that data back into
the network. Some huge deployments are going on at the moment in collaboration
with Hewlett Packard Enterprise (HPE) using
the HPE Apollo platform
to deploy high-density storage systems that remove the need to ship a set-top
box with a hard drive in it.
HPE Rethinks
And Use IT
Now, what are the advantages of that? Everybody thinks
it’s costly, so you’ve taken the hard drive out, you have the storage in the
network, and that’s clearly one element. But actually if you talk to any
operator, their biggest cause of subscriber churn is when somebody’s set-top
box fails and they lose their personalized recordings.
The
personal connection you had with your service isn’t there any longer. It’s a
lot easier to then look at competing services. So if that content is in the
network, then clearly you don’t have that churn issue. Not only can you access your
content from any mobile device, it’s protected and it will always be with you.
Taking the CDN private
Gardner: For the past few decades, part of
the solution to this problem was to employ a content delivery network (CDN) and
use that in a variety of ways. It started with web pages and the downloading of
flat graphic files. Now that's extended into all sorts of objects and content. Are
we going to do away with the CDN? Are we going to refactor it, is it going to
evolve? How does that pan out over the next decade?
Larbey: The CDN will still exist. That
still becomes the key way of optimizing video delivery -- but it changes. If
you go back 10 years, the only CDNs available were CDNs in the Internet. So it
was a shared service, you bought capacity on the shared service.
Even today that's how a lot of video from the content owners and broadcasters is streamed. For the past seven years, we have been taking that technology and deploying it in private network -- with both telcos and cable operators -- so they can have their own private CDN, and there are a lot of advantages to having your own private CDN.
You get complete control of the roadmap. You can start to introduce advanced features such as targeted ad insertion, blackout, and features like that to generate more revenue. You have complete control over the quality of experience, which you don't if you outsource to a shared service.
Even today that's how a lot of video from the content owners and broadcasters is streamed. For the past seven years, we have been taking that technology and deploying it in private network -- with both telcos and cable operators -- so they can have their own private CDN, and there are a lot of advantages to having your own private CDN.
You get complete control of the roadmap. You can start to introduce advanced features such as targeted ad insertion, blackout, and features like that to generate more revenue. You have complete control over the quality of experience, which you don't if you outsource to a shared service.
There
are a lot of advantages to having your own private CDN. You have
complete control over the quality of experience which you don't if you
outsource to a shared service.
What we’re
seeing now is both the programmers and broadcasters taking an interest in that private
CDN because they want the control. Video is their business, so the quality they
deliver is even more important to them. We’re seeing a lot of the programmers and
broadcasters starting to look at adopting the private CDN model as well.
The
challenge is how do you build that? You have to build for peak. Peak is
generally driven by live sporting events and one-off news events. So that leaves
you with a lot of capacity that’s sitting idle a lot of the time. With cloud and
orchestration, we have solved that technically -- we can add servers in very quickly,
we can take them out very quickly, react to the traffic demands and we can
technically move things around.
But the commercial
model has lagged behind. So we have been working with HPE Financial Services to understand how we
can innovate on that commercial model as well and get that flexibility -- not
just from an IT perspective, but also from a commercial perspective.
Gardner: Tell me about Private CDN technology. Is that
a Nokia product? Tell us about your business unit and the commercial models.
Larbey: We basically help as a business unit.
Anyone who has content -- be that broadcasters or programmers – they pay the operators
to stream the content over IP, and to launch new services. We have a product focused
on video networking: How to optimize a video, how it’s delivered, how it’s
streamed, and how it’s personalized.
It can be a private CDN product, which we have deployed for the last seven years, and we have a cloud digital video recorder (DVR) product, which is all about moving the storage capacity into the network. We also have a systems integration part, which brings a lot of technology together and allows operators to combine vendors and partners from the ecosystem into a complete end-to-end solution.
HPE Rethinks
And Use IT
Gardner: With HPE being a major supplier for
a lot of the hardware and infrastructure, how does the new cost model change
from the old model of pay up-front?
Flexible financial formats
Larbey: I would not classify HPE as a supplier;
I think they are our partner. We work very closely together. We use HPE ProLiant DL380 Gen9 Servers, the HPE Apollo platform, and the HPE Moonshot platform,
which are, as you know, world-leading compute-storage platforms that deliver
these services cost-effectively. We have had a long-term technical relationship.
We are now
moving toward how we advance the commercial relationship. We are working with
the HPE Financial Services team to look
at how we can get additional flexibility. There are a lot of pay-as-you-go-type
financial IT models that have been in existence for some time -- but these don’t
necessarily work for my applications from a financial perspective.
Our goal is to use 100
percent of the storage all of the time to maximize the cache hit-rate.
In the private CDN and the video applications, our goal is to use 100 percent of the storage all of the time to maximize the cache hit-rate. With the traditional IT payment model for storage, my application fundamentally breaks that. So having a partner like HPE that was flexible and could understand the application is really important.
We also
needed flexibility of compute scaling. We needed to be able to deploy for the
peak, but not pay for that peak at all times. That’s easy from the software technology
side, but we needed it from the commercial side as well.
And thirdly,
we have been trying to enter a new market and be focused on the programmers and
broadcasters, which is not our traditional segment. We have been deploying our CDN
to the largest telcos and cable operators in the world, but now, selling to
that programmers and broadcasters segment -- they are used to buying a service from
the Internet and they work in a different way and they have different
requirements.
So we
needed a financial model that allowed us to address that, but also a partner
who would take some of the risk, too, because we didn’t know if it was going to
be successful. Thankfully it has, and we have grown incredibly well, but it was
a risk at the start. Finding a partner like HPE
Financial Services who could share some of that risk was really important.
Gardner: These video delivery organizations
are increasingly operating on subscription basis, so they would like to have
their costs be incurred on a similar basis, so it all makes sense across the
services ecosystem.
Our tolerance just doesn't exist anymore for buffering and we demand and expect the highest-quality video.
Larbey: Yes, absolutely. That is becoming more and more important. If you go back to the very first the Internet video, you watched of a cat falling off a chair on YouTube. It didn’t matter if it was buffering, that wasn't relevant. Now, our tolerance just doesn’t exist anymore for buffering and we demand and expect the highest-quality video.
If TV in 2027
is going to be purely IP, then clearly that has to deliver exactly the same quality
of experience as the broadcasting technologies. And that creates challenges. The
biggest obvious example is if you go to any IP TV operator and look at their
streamed video channel that is live versus the one on broadcast, there is a big
delay.
So there is a lag between the live event and what you are seeing on your IP stream, which is 30 to 40 seconds. If you are in an apartment block, watching a live sporting event, and your neighbor sees it 30 to 40 seconds before you, that creates a big issue. A lot of the innovations we’re now doing with streaming technologies are to deliver that same broadcast experience.
HPE Rethinks
And Use IT
Gardner: We now also have to think about 4K resolution,
the intelligent edge, no latency, and all with managed costs. Fortunately at
this time HPE is also working on a lot of edge technologies, like Edgeline
and Universal
IoT, and so forth. There’s a lot more technology being driven to the edge
for storage, for large memory processing, and so forth. How are these advances
affecting your organization?
Optimal edge: functionality and storage
Larbey: There are two elements. The compute,
the edge, is absolutely critical. We are going to move all the intelligence
into the network, and clearly you need to reduce the latency, and you need to
able to scale that functionality. This functionality was scaled in millions of households,
and now it has to be done in the network. The only way you can effectively
build the network to handle that scale is to put as much functionality as you
can at the edge of the network.
The HPE
platforms will allow you to deploy that computer storage deep into the network,
and they are absolutely critical for our success. We will run our CDN, our ad
insertion, and all that capability as deeply into the network as an operator
wants to go -- and certainly the deeper, the better.
The other
thing we try to optimize all of the time is storage. One of the challenges with
network-based recording -- especially in the US due to the content-use regulations
compliance -- is that you have to store a copy per user. If, for example, both
of us record the same program, there are two versions of that program in the
cloud. That’s clearly very inefficient.
The
question is how do you optimize that, and also support just-in-time transcoding
techniques that have been talked about for some time. That would create the
right quality of bitrate on the fly, so you don’t have to store all the
different formats. It would dramatically reduce storage costs.
The
challenge has always been that the computing processing units (CPUs) needed to
do that, and that’s where HPE and the Moonshot platform, which has great
compute density, come in. We have the Intel media library for doing the
transcoding. It’s a really nice storage platform. But we still wanted to get
even more out of it, so at our Bell Labs research facility we developed a
capability called skim storage, which for a slight increase in storage, allows
us to double the number of transcodes we can do on a single CPU.
That approach
takes a really, really efficient hardware
platform with nice technology and doubles the density we can get from it --
and that’s a big change for the business case.
Gardner: It’s astonishing to think that that
much encoding would need to happen on the fly for a mass market; that’s a
tremendous amount of compute, and an intense compute requirement.
Content popularity
Larbey: Absolutely, and you have to be intelligent about it. At the end of the day, human behavior works in our favor. If you look at most programs that people record, if they do not watch within the first seven days, they are probably not going to watch that recording. That content in particular then can be optimized from a storage perspective. You still need the ability to recreate it on the fly, but it improves the scale model.
Gardner: So the more intelligent you can be
about what the users’ behavior and/or their use patterns, the more efficient
you can be. Intelligence seems to be the real key here.
Larbey: Yes, we have a number of algorithms
even within the CDN itself today that predict content popularity. We want to
maximize the disk usage. We want the popular content on the disk, so what’s the
point of us deleting a piece of a popular content just because a piece of long-tail
content has been requested. We do a lot of algorithms looking at and trying to
predict the content popularity so that we can make sure we are optimizing the
hardware platform accordingly.
Gardner: Perhaps we can deepen our knowledge
about this all through some examples. Do have some examples that demonstrate
how your clients and customers are taking these new technologies and making
better business decisions that help them in their cost structure -- but also
deliver a far better user experience?
In-house control
Larbey: One of our largest customers is Liberty Global, with a large number of
cable operators in a variety of countries across Europe. They were enhancing an
IP service. They started with an Internet-based CDN and that’s how they were
delivering their service. But recognizing the importance of gaining more
control over costs and the quality experience, they wanted to take that in-house
and put the content on a private CDN.
We worked
with them to deliver that technology. One of things that they noticed very
quickly, which I don’t think they were expecting, was a dramatic reduction in
the number of people calling in to complain because the stream had stopped or buffered.
They enjoyed a big decrease in call-center calls as soon as they switched on
our new CDN technology, which is quite an interesting use-case benefit.
When they deployed a private CDN, they reached costs payback in
less than 12 months.
One of our
customers in Canada, TELUS, when
they deployed a private CDN, they reached costs payback in less than 12 months in
terms of both the network savings and the Internet CDN costs savings.
Gardner: Before we close out, perhaps a look
to the future and thinking about some of the requirements on business models as
we leverage edge intelligence. What about personalization services, or even
inserting ads in different ways? Can there be more of a two-way relationship, or
a one-to-one interaction with the end consumers? What are the increased
benefits from that high-performing, high-efficiency edge architecture?
VR vision and beyond
Larbey: All of that generates more traffic
-- moving from standard-definition to high-definition to 4K, to beyond 4K -- it
all generates more network traffic. You then take into account a 360-degree-video
capability and virtual reality (VR) services, which is a focus for Nokia with
our Ozo camera, and it’s clear that the
data is just going to explode.
So being
able to optimize, and continue to optimize that, in terms of new codec
technology and new streaming technologies -- to be able to constrain the growth
of video demands on the network – is essential, otherwise the traffic would
just explode.
There is
lot of innovation going on to optimize the content experience. People may not want
to watch all their TV through VR headsets. That may not become the way you want
to watch the latest episode of Game of Thrones. However, maybe there will be a
uniquely created piece of content that’s an add-on in 360, and the real serious
fans can go and look for it. I think we will see new types of content being
created to address these different use-cases.
Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript or download
a copy. Sponsor: Hewlett Packard Enterprise.
You may also be
interested in:
