Layer123's note - this is part 2 of 2, you can read the 'Virtualisation' and 'New Models' sections here.
Dr. Mark H Mortensen, Principal Analyst at ACG Research, spent three days at the Layer123 Reunion: Intelligent Network Automation Congress in April - and here is his summary of what the speakers and delegates thought about some key issues in our industry.
I attended the excellent Layer 123 Reunion conference in Madrid, Spain from 27-28 April. Here I discuss the major topics and generally agreed-to points I saw in the conferences – and MYTAKE personal opinions.
This is meant to document my understanding of what I heard – If you were there, please let me know if you disagree with my interpretations of the proceedings. And everyone is invited to comment on my opinions on LinkedIn, where I will be posting this document broken into the four topical areas at www.linkedin.com/in/markhmortensen .
Network automation is a critical need.
- We are converging on an industry vision that has been described by many SDOs and forums in slightly different ways, but all are about the same. The networks must take care of themselves, without human intervention. ETSI (ZTSN) and TM Forum (Autonomous Networks) are the broadest and best articulated.
- The problem of automating services operations is basically top-down, using cross-domain orchestrators coordinating domain controllers.
- The problem of automating network operations is basically bottom-up, using task and process automation within domains and making operations within a domain as autonomous as practical, but managed by the upper layers via a rich API.
- We will need some major advances in technology to make network automation be what we need it to be:
- Visibility and control of all of these disaggregated network functions running on a distributed processing platform and moving around as the needs change.
- And the same for other applications owned by the CSPs and the customers.
- Explanation AI (XAI) capable of telling us why it came to its conclusion so we can learn to trust the bots.
- Digital twins to allow us to not only understand what is happening in the network, but also to try out new services, new configurations, new management software, and new policies in general.
- Data lakes that contain the terabytes of information coming from the network, making them available to the analytics applications sipping from the lakes.
- A common inventory view of the infrastructure used by all.
- More common data model to underpin the various APIs.
- MYTAKE: The best articulation of this vision is in the TMForum Autonomous Networks project. It deserves to be, and will be, the center of gravity for the multitude of projects and programs to advance the concept.
Software Defined Networking is now a given.
- Most network elements (physical and virtual) being deployed now have NETCONF interfaces and support YANG models, the basic requirements of Software Defined Networking. They are adopting the standard APIs from the likes of the TM Forum, IETF, Broad Band Forum, 3GPP and other SDOs.
- MYTAKE: Only a little over half of the elements being deployed are actually under SDN control today, scheduled to rise to about 75% by 2027. Work is necessary for the Domain Controllers to really work well and the cross-domain controllers to coordinate across domains. I believe most domain controllers will come from the network element vendors themselves (except where the CNFs are vertically disaggregated, where it is not clear who will do the DC function). And the CDOs will predominantly come from independent software vendors.
Intent-based interfaces are the future, but hard to do.
- The translation of customer requirements to product requirements is time-consuming, expensive, and fraught with mistakes. Then the translation from product requirements into the detailed network resource requirements are nearly as difficult.
- Customers, conditioned by the hyperscalers, want it NOW.
- We don’t really understand yet what a real Intent-Based Interface is, other than a general description – a way of completely specifying what the customer wants from their service, without the need for any additional “common sense” information.
- Work is underway in many vendors and SDOs to work this issue.
- MYTAKE: This is going to be really hard. We “punted” in the 3GPP slicing specs that use parametrized “templates” for the various sliced services, rather than an intent-based approach.Some CSPs have gotten around the problem by creating pre-packages of vertical-specific services (e.g. gold or silver versions of a service for connecting consumers to a gaming platform with low latency). But the formal specification of general intent-based interfaces will be very hard. We will still be talking about it in five, or maybe even 10 years, although we may be able to do a good job on some of the simpler services. The interface between the services and resources is needed to give the isolation we need in the AN for truly autonomous domains. The customer interface is the harder problem and, although important to the business to support a highly generalized, sophisticated NaaS service interface to the customer, is less important to the network automation challenge.
Network security is ever more critical but changing rapidly.
- The shift from the definition of a security perimeter being location-based to being user/device-based means new requirements. SASE in SD-WAN is an example of this.
- SASE will have to mutate significantly in the next few years.
- Security of edge platforms, the third-party software running on edge platforms, CNFs, their management, the AIs that manage them, and all the supply chains of these components will have to be strengthened.
- IT-Network-Security people need to work together.
- We’ve already seen the need to combine the CTO and CIO organizations with virtualization, creating CTIOs.
- Security needs to be better integrated into the plans, too, not just added-on later. This will require more joint planning among the specialties. Combining the CTO, CIO and Security organizations into one organization sounds like a solution, except these combined organizations cannot be just under one head, more is needed.
- Personnel rotations and cross-organizational project teams were suggested by the CSPs as practical solutions.
- MYTAKE: I remember when operations was added-on to the network and service plans at the last minute. Although that is still the case in many CSPs, the progressive ones have realized the need for the OSS/BSS and overall operations plans to be done as part of the network and service evolution planning. We will see the same evolution on security, but, hopefully, faster.
Edge computing is a great opportunity.
- Although there are many definitions of the “edge,” CSP-owned edge computing platforms on the customer premises, at the local central office, and maybe regionally will be an important part of their future.
- Many of the applications running on the MEC infrastructure will be for internal CSP use (e.g., non-Real Time RIC functions in a disaggregated O-RAN, virtual cell-site routers, or CSP-provided virtual firewalls for a customer).
- MYTAKE: I have been a skeptic for a long time about how much MEC there will be hosting low-latency services (believed by most to be the major third-party use case – there are others). The push by the hyperscalers to help colonize the edge tells me I may be wrong. And the growing list of internal CSP uses may be enough to make it a major part of the network of the future. But it is clear, distributed computing will win, at least for a while – it always does, after the pendulum swing towards consolidation, as we just went through.
The Next Generation Network requirements will come from Industry 4.0 and the Metaverse.
- The requirements of the next generation network will be set for:
- Enterprises by the Intelligent Composable Fabric (ICF) that will be used to construct the digital enablement systems for enterprises for their Industry 4.0 digitalization
- Consumer by whatever the Metaverse becomes (of course, there is some crossover to enterprise here, too) that will bring virtual and augmented reality to human communications and entertainment.
- The nextgen network requirements are not a new set but are reprioritized.
- MYTAKE: Key challenges are:
- Defining and deploying a true NaaS infrastructure
- Defining and scaling autonomous networks
- Electronically bonding CSPs for ordering and continued operations across geographical boundaries
- Building a distributed computing infrastructure
- Building Industry 4.0 and Metaverse ecosystems.
CSPs could do this – we’ll see what part they decide to play.
Comments are welcomed.