Recently, Ericsson Digital released an amazing report on Edge Computing and 5G. In it, they explained how distributed cloud computing is paving the way for the future of network communications. They also advised telecommunications operators, not to wait for 5G but they should build tomorrow’s networks today! Given the cost, effort and focus such a commitment would demand, I was anxious to learn the reasoning behind such a bold statement. My curiosity was answered when they gave me the opportunity to interview Mats Johansson, Senior Marketing Dir, IP & Cloud Ericsson.
Kevin: Mats, thank you very much for taking the time to discuss your recent report, “Edge Computing and 5G” with me!
Mats: I’m happy to speak with you today Kevin.
Kevin: Thanks. The report really piqued my curiosity.
Given that the current global buildout of 5G networks is top of mind for just
about everyone today, the focus of your report seems to be edge computing and
distributed clouds. Could you please explain the difference between cloud
cloud and edge computing? What do these terms mean?
Mats: The network service providers are developing
cloud platforms for just about everything in their core. Today more than 90% of
that core is composed of physical node which can’t support rapidly increasing
traffic. This traffic increase and resulting flat revenues is driving the need
for more efficiency through the adoption of automation. They are turning the
network into an “android phone”. Essentially, they need to deploy different
types of hardware, an operating system, a virtualization layer, applications
and an app store.
Kevin: I love that analogy. If this
change is as simple as building a supersized smartphone, what’s the problem?
Mats: The main difference is that operators can’t afford
to have any down time during updates! When you upgrade your smartphone, you
can’t use it for anything. That can’t happen when the service providers upgrade
their networks. This alone adds a great amount of complexity. In addition to
that, there are multiple generations of technology in service. They have 2G, 3G,
4G, and 5G deployed. Physical nodes will only slowly disappear as cloud
computing is adopted.
Distributed cloud is needed to enable short network latency
when running advanced applications (e.g. gaming, self-driving cars, augmented
reality, factory automation), enable the caching of data (e.g. delivery of high
definition Netflix, HBO, and other video on demand services) near end users which
is known as “transport off-load”, process off-loading to local devices to avoid
a dependency on centralized datacenters for complex computation services ( e.g.
facial recognition for access control) , to expose advanced network
capabilities to application developers.
Edge Computing only the compute part of this complex
picture. It only addresses process offloading.
So distributed cloud requires the entire network
infrastructure which is especially important for application developers. They
need API (application programming interface) access to the advanced
network-based services without worrying about where these services are located.
The telecom network itself needs to be able to place the advance service
workloads where they need to be.
Kevin: You’re basically abstracting all the telecommunications
stuff from the application developer. All they need to do is call the service
with an API.
Mats: That’s right. They would simply enter
application requirements as API parameters when calling the service.
Kevin: This seems like quite a big change for the operator’s
customers. In the past they have needed to deal with telephone service
providers, internet service providers and now distributed cloud service
providers. Will this never-ending increased complexity for the consumer ever
Mats: There are two types of customers to consider
here. The general consumer and the enterprise customers. For the general
consumer there is no change. For the enterprise application developer this will
be no more complex than developing an application for any app store. This will,
however, expose opportunities that they have never had before. All the
additional complexity is taken on by the network service provider, not their
Kevin: This looks like a huge business opportunity
for communications service providers. Can these network operators step up to
this challenge before the cloud service providers capture the opportunity?
Mats: This is a battle that will play out in the
marketplace but telcos may have an advantage. When considering datacenter real
estate for example, telcos have more real estate than cloud service providers.
This is playing out today in Germany where there is heavy competition between
telcos, cloud service providers and large corporate enterprises. Cloud service
providers will also need to invest heavily in this area if they want to
There is also a technical challenge for these web
players. They will always have a
dependency on telcos for their IP connections which is always routed through
their core network. Telcos can more easily manage edge network IP connections
which could give web players much more difficulty. For example, when you travel
from the US to Europe, your smartphone IP connection needs to “trombone” back
and forth back to your US home network.
This is due to network management through the user plane function. This
makes roaming service very difficult and this issue cannot be easily solved by
the web players. While web players may build some telecommunications sites,
they will also partner with network operators.
Kevin: So, it looks like the next phase of “Cloud
Wars” will inevitably include network operators. How do telco companies address
the cultural and operational challenges that this will bring? Things like
adopting NFV (network function virtualization), SDN (software defined networks)
and software development using the CI/CD (Continuous Integration/Continuous
Deployment) model are old hat for the web players.
Mats: Operators have a significant internal political
battle around protecting existing individual domains. Today the department responsible for the
packet core runs and manages the entire technology stack and the department responsible
for IMS/voice runs and manages that stack. This makes current deployments very
siloed. Using NFV and 5G leads to a single platform for all applications and a
single network for all services. A significant number of operators don’t have a
standardized technology approach across their many sites. Instead of a
unifiedcloud platform, many operators have multiple vertical cloud stacks. On
one site, for instance, they could have an IMS stack from Ericsson and a packet
core from Affirmed while another site could be using Red Hat for IMS, an Ericsson
Openstack distribution (CE) for their core and Huawei for IoT (Internet of
To ideally transition to a world that includes network
slicing and 5G, operators need a horizontally integrated single cloud platform
for all applications. Then they can enjoy the benefits and efficiencies of
cloud and would then be able to deploy a single network for all services. That’s
the aspiration of 5G! One network that can simultaneously support “elephant
flows” like enterprise datacenter backup, low latency services like remote
surgery or self-driving cars and the really small flows like smart metering and
sensors for roads and agricultural businesses. If you don’t build one network,
you would need to build dedicated networks for each of these solutions which is
much more expensive. A single network that leverages network slicing also needs
network virtualization (NFV, SDN) and the distributed cloud. So, the whole
machinery fits together!
Kevin: How can Ericsson help with this transition?
Mats: We can help operators work through the
operational transformation with our
consulting services. Our consultants have learned how to do this by working
with leading operators like Swisscom that have already done this. They
reorganized their internal teams so that now, one team operates the cloud
platform and other team manage the applications. The teams have quarterly joint
planning meetings for DevOp type upgrades and quick meetings between the
quarterly get togethers for smaller upgrades. Along side this, you need to
deploy automation across all layers of the stack. Up until now, upgrades have
been implemented using manual procedures which are typically yearly upgrades.
The new model can deliver monthly or even daily upgrades with cloud native
deployments at the microservice level.
Kevin: How can Ericsson products help with this?
Mats: We can provide all the NFV stack
layers. For the cloud infrastructure we have NFVi, the virtual
infrastructure manager, SDN for networking within the datacenter and software
defined infrastructure. In collaboration
with Intel we have the SDI (software defined infrastructure) solution that gives
you the ability to optimize a pool of resources without needing to buy new
datacenter hardware if you run out (e.g. storage). This can be used across
multiple hardware brands including HP, Dell and Ericsson.
Kevin: To me it sounds like Ericsson can help with
both the cultural change and the automation.
Mats: Yes. Going back to the smart phone analogy,
there is no point downloading the apps without the App Store and Android operating
system layer! Now, operators are struggling to get the hardware and the operating
system layer to work. Therefore, vertical stacks seem easier to them. Network
operators use a hardware server platform, some virtual machines and a few VNFs
but this can’t easily scale to a large 5G network that can use network slicing!
Some Ericsson network operator customers can offer new opportunities
to their enterprise customers today. Those customers can order services via a
services portal (firewalls, SDWAN solution) and use third party VNF’s that have
already been certified.
Kevin: With all these needed changes and new business
opportunities on the horizon, how will network operator morph their current operational
models into becoming distributed cloud operators and edge solution providers?
Mats: That’s a very good question that doesn’t have a
simple answer. The challenge is that we will see two models. One is the horizontal
cloud paired with an in parallel operation transformation. Full-fledged
orchestration comes when the basic NFVI platform works. Ericsson advocates
buying a pre-integrated stack for the NFVI as we have seen self-integration as
The second model is through the adoption of vertical
solutions. In this option you don’t have to reorganize each internal operating
domain. Each remains in its own Kingdom which makes this an attractive option
for solving the political problems. It will be interesting though to see whether
those following this path will need to eventually transition to a more
horizontal cloud as it may be very hard to manage services via network slicing.
This model will likely be harder and more expensive.
Kevin: Given these two models, how will this distributed cloud future affect competition between network operators?
Mats: Those who build orchestrated cloud and get it
working, will be in a better position to capture the opportunities. They will
enjoy lower operational expenses and have an advantage in capturing local
Kevin: That is certainly an important point. For my
final question Mats, what would be your advice to a telco CEO that may be
reading this post?
Mats: I would tell them to start their cloudification
transformation now. The journey from start to running is typically 3-4 years.
You need to engage and understand what the implications really are and what business
opportunities you may be missing. Operators
that don’t build an edge computing capability will miss out on 25% of the new
5G use cases! With that in mind, you should start now with getting an
understanding of what transformation for your company is going to take. Hire
the right people and build a horizontal orchestration platform if you want to
be a competitive 5G player.
Kevin: Thank you for such an insightful conversation!
Mats: My pleasure Kevin.
This post was sponsored by Ericsson Digital