The recent Telco Cloud 2022 conference brought to the fore many of the key challenges that CSPs face with digital transformation.

Omdia view


Anyone familiar with telco transformation understands that the process is never easy. The recent Telco Cloud 2022 conference (part of the Network X event, alongside Broadband World Forum and 5G World) brought to the fore many of the key challenges that communications service providers (CSPs) face.

In public presentations and private discussions, CSPs lamented the complex and dynamic environment they are trying to transform. There are a lot of moving parts, and every issue is related but also independent of the others.

Some of the key areas covered during the event were automation, data analytics, and (of course) cloudification. One common theme was the necessity, and difficulty, of convincing upper management and front-line operations staff about the benefits of deploying each of these technologies.

Data analytics

For many CSPs, the thought of sorting through their massive troves of data is daunting. They know that their data is a mess. They also know that it is valuable information and that the cloud can play a key role in applying analytics to extract that value. It was clear from the questions asked during sessions that CSPs are looking for help to better understand how to do so.

Part of the challenge comes from the fact that data usually resides in siloes in different organizations. Getting finance, network operations, and marketing, for example, to make their data available to others is challenging at best and impossible at worst. Network operations teams face the additional challenge of extracting data from proprietary element management systems (EMS). And even if they were convinced to share their data, harmonizing on a standard taxonomy to be used in all situations—and ensuring the relevance, completeness, and accuracy of the data—would still loom large. CSPs cannot begin to leverage analytics until the data is clean, which is why many remain stuck on the sidelines.

Another organizational challenge that outsiders may not appreciate is that many CSPs place their data science teams within the R&D organization, where what they work on is determined by the business goals that are set. In private conversations, CSPs expressed frustration in getting access to these data scientists to work on network-related initiatives.

Outside suppliers could assist CSPs with their data-led transformations, but many of them are unwilling to partner with suppliers. They know that they have value in their data and do not want to share it until they better understand it. This means they mostly work in-house and, as mentioned earlier, in siloes. One thing everyone agreed on was the critical need for senior-level sponsorship and support for companywide data transformation.


Despite most CSPs ceasing to offer their own public cloud services, they continue to build private clouds and use the public clouds of the hyperscalers. Existing clouds, private and public, are sufficient to host some enterprise IT functions, but few are robust enough to host network functions. CSPs rely heavily on their network function vendors to bring their own cloud capabilities. Making the matter worse is that vendors select the stack that optimizes the performance of a given network function, meaning there could be different stacks even from the same vendor. Thus, in a multi-vendor network, CSPs have to manage several instances of Kubernetes, OpenStack, etc. This makes it hard to standardize on lifecycle management processes, which makes it very difficult to orchestrate end-to-end services. Moreover, this complexity means opex could go up because CSPs cannot realize any economies of scale.

CSPs also spoke about some of their challenges in running workloads on hyperscalers’ public clouds. In addition to the aforementioned technical challenges, CSPs face commercial issues that hinder them from using the cloud for analytics. It is not that storage costs are so high, but rather, that it costs them every time they want to take data out. For this reason, CSPs may choose to do more processing on-premises or at the edge and then move just the results of the analysis to the cloud.


Automation is a broad topic that comes up in many contexts and can mean many different things. Today, CSPs can use basic scripting to automate some equipment configurations, and inventory systems can auto-discover new equipment that has been deployed. In the future, autonomous networks could self-provision, self-heal, and self-optimize without any human intervention. Most CSPs are closer to the first example than the latter. Nearly all seem to understand that they need to automate more of their processes—especially with the advent of 5G standalone (SA)—but few know how to move toward higher levels of automation, and they do not want to make mistakes.

Some of the reasons posited as to why CSPs are not moving faster on automation include the following:

  • Employees do not have the skill sets to implement and manage automated systems.
  • The lack of standardization across different hardware and software vendors leads to interoperability issues.
  • Employees resist because they are concerned automation will take over their jobs and make them redundant.
  • Network operations staff do not yet trust the output of automation systems and are unwilling to hand over control.

However, CSPs expressed optimism that most (if not all) of these challenges could be overcome in time. What was clear, though, is that they are not going to move any more quickly than they think is prudent, given their mission to sustain network and service availability.



Roz Roseboro, Principal Analyst, Service Provider Transformation

Christoforos Sarantopoulos, Senior Analyst, Service Provider Transformation