Updated: Mar 8
In the first part of our MWC series, we had opined at some length about the telco “burning platform” and the tough choices that they need to make to survive and thrive. These tough choices are not new…in many ways, they have become a recurring theme for telcos. This intractable problem has seen many attempts to break through but relative success to date. Most telco efforts to tackle this intractable problem have come up short for a variety of reasons, many for having too narrow a field of vision, many for not having enough support across the operator community. However, there is an argument to be made that they failed because they were simply not bold enough.
This intractable problem is reminiscent of the story about the Gordian Knot. In the ancient kingdom of Phrygia (in modern day Anatolia in Turkiye), the story goes that the Phrygians chose Gordias to be their King as he rode into the city in an ox-cart and fulfilled a prophecy. The ox-cart was then tied to a post with an intricate “knot” as a dedication to the Phygian god Sabazios. While the local legends foretold that the man who was able to untie the knot would become the “ruler of all Asia”, no one had ever come close until Alexander the Great arrived in Phrygia in the 4th century BC. Alexander, a man with clear ambitions to head East, had great incentive to untie the knot and fulfil the prophecy and he did succeed. However there are two versions of the story. In one, Alexander simply didn’t bother with tracing the knots to their sources, he just sliced through them with his sword. In the other version, he unhitched the knot from the post to reveal the source point of the knots. But the truth of how he did it is almost irrelevant. The story would suggest that the best way to solve an intractable problem is to ignore the premise of the original problem to be solved.
How is this story relevant to telcos, you ask? No analogy is perfect but we would like to point to the fact that one of the biggest reasons why telcos have not been able to solve the intractable problem is that they have not been able to the Gordian Knot that is their infrastructure. It’s not that attempts haven’t been made in the past. NFV anyone? SDN? Cloud-native? Open RAN? Disaggregated networks? BSS/OSS 2.0? None of these initiatives have failed per se but none of them have delivered telcos to the promised land. So let’s update the statement a little. One of the biggest reasons why telcos have struggled to deal with the intractable problem is that they have not solved the Gordian Knot that is their underlying infrastructure as all attempts have been incremental and ultimately didn’t go far enough. Why? Because all of these telcos had to protect their legacy businesses, or at the very least, not disrupt them as they were cash cows. However, this incremental, conservative approach is not sustainable in that telcos will not be able to monetize their 5G investments without fundamental changes to the way they architect their networks by flattening and automating them.
At MWC 2023, we had some discussions that would suggest that telcos may finally be on the verge of untangling the Gordian knot. While it is impossible to cover all the emerging scenarios, we will highlight a few conversations we had at the show and potential approaches.
Segment routing for network slicing
One of the key issues we have identified is that many of 5G’s best features and concepts are currently difficult to implement because the routing of data packets through the network continues to be challenged by the increasing diversity of nodes and segments to be navigated for the optimal routing of IP packets. Without true end to end visibility and optimized routing, services like “network slicing” will never take off. As things stand, today’s network infrastructure is struggling to meet these new demands due to the dichotomy in telecommunications and networking domains, each governed by different sets of standards and incompatible technologies. But bridging the two domains has proved challenging, with a multitude of protocols and software used to bridge and interconnect the various nodes straddling the two domains. MPLS has acted as the bridge between these two domains, but it will not scale. With the advent of 5G, the complexity and fragmentation in the transport layer of the network has become a critical issue to solve.
One way forward to realize the potential of network slicing is to explore segment routing over IPv6 (Srv6), which is well suited for 5G transport and backhaul as it can reduce the number of protocols running on IP/MPLS networks, simplifying O&M complexity. At MWC23, we came across a great example of the shift towards SRv6 when we met Arrcus (providing a state-of-the-art network operating system called ArcOS) in partnership with Softbank, the Japanese mobile operator and Intel (providing the silicon and server hardware). Softbank has pioneered a new architecture called Mobile User Plane (MUP) that uses SRv6 as the bridge between the 5G and IP networks. MUP will allow mobile operators to utilize commercial IP routers and switches as opposed to customized hardware. By using SRv6, MUP will also simplify the transport layer significantly by eliminating the cumbersome protocols like RSVP-TE that have to be used to stitch together a seamless path through the network. The future of network slicing may just be closer than we currently believe.
Many of our readers will say to my face that we just made up “interworking”. We only wish we had. We are glad that there are serious people tackling these problems in the network. Every telco at one time or another has had to deal with this issue. Interworking ensures communication (i.e. signaling) between network nodes works when the nodes connecting are from a different technology (e.g. 3G and 4G). This is vital to make mobile services operate across an increasing number of connected nodes that are operating on different generations of communications technologies. There are still several examples of mobile networks that operate multiple generations, from 2G all the way through 5G. While 2G/3G network sunsets have become increasingly commonplace, the challenge remains, to maintain legacy systems while investing in 5G. This results most often in frustrating operators due to increased complexity as well as the mounting operating costs. Case in point is the need for operators to ensure roaming fall back scenarios on 2G or 3G, potentially for years to come.
One of the foundations within the depths of the telecoms network is signalling. When signaling is interrupted networks simply grinds to a halt. The huge problem for telcos is that with every subsequent generation of cellular technology deployed, their innate conservatism and need to depreciate investments in older network technologies has meant that making services work across signaling variants has become incredibly complicated. Therefore operators with such products have to rely on the availability of individual infrastructure vendors to stitch together layers of (incompatible) network elements.
Now with 5G, network functions are moving into the cloud, but this won’t address the fundamental challenge. Industry giants like Oracle and other vendors herald the era of Cloud-native core network functions. But such microservices enabled network functions come with a hefty price tag for each instance running in the cloud. Moreover cloud nativeness does nothing to solve interworking. For example to address data stored on former technology systems to prevent duplication of the same data between different technologies. Yet another Gordian Knot. Our meeting with BroadForward at MWC23 led us to believe that there are solutions and approaches to these problems. BroadForward (a GLOMO nominee), is a small company of dedicated telecom professionals out of the Netherlands, and with intelligent core network software solutions. Each of these are based on the common converged software architecture called BFX. It comes with 2G, 3G, 4G and 5G interworking capabilities by design for each of their network functions. Their solution greatly reduces complexity while bridging multiple generations of network technology as well as being able to run in the cloud on minimal instances and hardware resources.
Orchestration through the telco stack
One of the emerging areas that requires increasing attention from telcos is at the edge of the network. Increasingly, network resources are getting nudged or dragged (kicking and screaming), depending on your perspective, to the edge of the network to cater to new use cases and application workloads of enterprises and even consumer services. These new workloads can cut across industrial IoT to cloud gaming, to autonomous vehicles and many more. Some of these workloads will be “processed” on-premise at the enterprise site with edge computing resources, while some will go back to the cloud, private or public.
But today, edge networks are neither performant nor are they capable of handling the increasing demands. For telcos, here is another intractable problem, another Gordian Knot. Do they invest in more network compute resources at the edge and how do they efficiently move orchestrate services and workloads across from the edge of the network through the stack to the underlying hardware and ultimately back to the cloud and/or core network. This needs to be done in a highly automated manner, and telcos are struggling with these demands today.
At MWC23, we met with EnterpriseWeb, a startup innovator who is working on breaking down the “lack of unified abstractions and common tooling that has been the roadblock to digital business transformation”. Occupying a very small footprint, the CloudNFV platform is an attempt to become a next generation, “no-code” OSS platform that allows orchestration up and down the stack in agile manner without expensive one-off solutions that create silos and complexity. At MWC23, EnterpriseWeb collaborated with Intel, Red Hat, Fortinet, Keysight Technologies, KX and Tech Mahindra on a 5G RAN Testbed that was designed to leverage software-defined infrastructure to expose standard interfaces across the stack. The orchestration engine would then “configure those physical and virtual components to drive continuous optimizations” driven by changing traffic patterns.
Modern Data Infrastructures
Previously, we have written about the need for telcos to modernize their data infrastructures so to be able to tap into the mountains of data streaming through their network. Monetization is not truly possible until the massive surges in data through the network are stored, processed and analyzed to generate real business insights. Unlocking the data that resides within telco networks is another intractable problem, another manifestation of the Gordian Knot. Unlocking this data will require fundamental changes in the data architecture deployed by telcos today, with three primary architectural elements that can be implemented.
The first of these is to implement “data lakehouses” to bring order the multiple data repositories spread through their networks, both in country and in the case of multinational operators, across countries. Second, telcos need a data fabric to access data stored in multiple repositories outside the telco network, in an enterprise’s on-premise datacenter, or in a private or public cloud. Third, while the data fabric can stitch together data assets that are spread across multiple clouds, Data from multiple layers of the stack are sometimes required for business queries within the telco, or even for use by third party enterprise clients and partners. A “data mesh” can connect all data assets within a data lakehouse or in multiple locations accessed through the data fabric, and wrap these with advanced tools and features that will elevate the data to business level insights.
At MWC23, we met with Cloudera, one of the companies working closely with telcos to modernize data architectures to meet with the hybrid, multi-cloud future. Cloudera works with operators across the world, though we have focused most on their Asian customers. A good example of efforts being made in this domain is the work done recently by Telkomsel, a leading operator in Indonesia. Faced with data generation of over 50 TB per day, Telkomsel began making investments, in partnership with Cloudera and Wipro, to migrate to a modern data architecture to cope with the increasing demands. The new architecture houses all Telkomsel internal data in a new data lake but also uses Apache NiFi as the data fabric to ingest first, second and even third-party data. The net result of this phase of investment has been significant reductions (nearly 60%) in infrastructure and solution delivery costs, as well as improved customer experience and faster app development cycles.
It has become customary for the telecoms industry to be half-glass empty. Constantly surrounded by declarations of impending gloom, the industry has not exactly redeemed itself with a narrow field of vision, bickering within the tent, and continuing misfires in execution. However, there remains enough innovation that provides the telco industry with the hope to unlock the Gordian Knots that have bedeviled the industry. We hope we have been able to reasonably highlight a few of these efforts and we will continue to research and analyze these developments. Our next post in the series on MWC23 will focus on private 5G. Stay tuned…