It has been more than a decade since this term has been floating around. Tons of ink have been used to describe, standardize and even code the TelcoCloud. Moreover, a growing business is being developed with old Telco actors, middle-aged Cloud actors and some newbies fighting for a piece of the cake.

I have been also around the term for six years now, first as TelcoCloud user and lately as TelcoCloud architect. And to be completely honest, I do not have a precise idea about what TelcoCloud means. So please do not ask me if a certain Cloud Platform can be considered as TelcoCloud. My answer will be that you need to reformulate your question. But let’s debug it.

Prior to the arrival of TelcoCloud platforms, the cloud industry already had a powerful technology offering. In a very simplistic way, these technical offerings can be classified and positioned using a two-axis plot with features&performance as X&Y coordinates.

Y-axis/Performance: This is the capability of the solution to host high intensive workloads, both in compute and packet processing, with excellent responsiveness. Techniques as cpu pinning, NUMA awareness, hugemem pages or data plane acceleration¹ are valued in this category. And of course the availability of high end performance servers and network fabric architectures are included.

X-axis/Features: It measures the service catalog available to deploy workloads like IaaS or CaaS for vitualization layer or SaaS for managed services. But also the tools the cloud offers to the customers to govern an orchestrate their deployment. From just the use of rich descriptors to describe and deploy your workloads as IaaC² , to advanced tools based on machine learning and artificial intelligence.

Using these two axis a simple quadrant can be drawn, being square one the traditional and in-decline physical hosting business. Besides that starting point, it is honest to admit there are two leagues: cloud on-premise built by IT departments of relevant companies operating generic business and the solutions developed by the big Cloud Providers also known as hyperscalers³.

Of course, it is not a black or white world and a handful of companies⁴ out of the hyperscalers selected club are positioning their cloud infrastructures in the Top Offering square and even challenging some of the hyperscaler’s business. But it does not change much the general picture.

But then, the Telco industry needed Cloud Technology to deploy their workloads also known as Network Functions. The question was: Is there a proper Cloud Platform? Proper to abandon the high efficient dedicated hardware traditionally used also known as baremetal and jump to the new world. The industry said no, there wasn’t, we need to invent a new one paradigm. And then the term TelcoCloud and other related aliases appeared.

Important projects were then initiated, both at private corporate level and in the standardization bodies⁵. Telco operators and telco equipment vendors drove a search for the proper environment to deploy the vital Network Functions that allow us to run our business, read techie posts on internet or hang out with friends during weekend. As mentioned before, the main assumption was that existing solutions cannot offer the needed performance and functionality. But was it really true? Let’s think about it using the previous simple diagram and try to place n it the current proposal implicit in TelcoCloud current solutions.

Regarding the performance axis, a lot has been said about the traffic intensity and quality of service needed in terms of bandwidth, packet loss, latency or jitter. Techniques to ensure these performance levels like CPU pinning or data plane acceleration have been considered a MUST for telco workloads. But do not be tricked, this techniques have been in the top offering of Cloud Technology for a decade⁶. And probably telco workloads have not been the first to use it, another industries like financial services have also demanded and used it for long.

The feature axis has been considered also crucial for Telco. The complexity of the infrastructure of a Network Service Provider is big, think of hundreds or thousands of interconnected nodes that are needed to be managed in perfect harmony. Then Orchestration came to light. On top of the execution environments a well defined Orchestration layer and a standardized operation model is highly appreciated. In fact, some standards like ETSI NFV are mainly focused on this, not in the performance axis. The concept of complex network services composed of several Network Functions and the interconnections between them, defined and managed by software, is pointed as a critical feature. But again, if we think of hyperscalers and their business, they are able to deal with the self-provisioning of hundreds of thousands customers that use a large catalog of services, most of them interconnected. It is not really new.

Therefore we can agree that the technology and tools were already there.

However, and to be fair, maybe there is a new field were the generic Cloud Industry have been challenged by the telco world. This is the so called far-edge. The capillarity in telco networks when it comes to the access planes like the radio network and others is huge. Deploying and using many point or presences is a feature that, although hyperscalers are exploring⁷, cannot be taken for granted.

But at the end, if we draw the technology proposal implicit in TelcoCloud we do not see, putting aside far-edge, anything to be invented.

Moreover, there is another aspect that is very often forgotten: the real requirements of the Network Functions. The Network Function world covers a wide spectrum of applications, from highly specific and performance demanding nodes like 5G UPF⁸ to more generic and less demanding like a DNS server. Do we need data plane accelerator techniques or complex orchestration to deploy a static set of centralized DNS servers? My answer is I don’t see it. The virtualization of such kind of “Network Functions” can be done using more inexpensive and less complex technologies like the ones behind enhanced traditional IT cloud on-premise.

At summary and going back to the initial question of what a TelcoCloud Platform is, I would say the one or the ones that serve well for the technical requirements and operational model of each specific Network Function or Service. As simple as that.

And to end this post, an important disclaimer: all previous insights are just under a strict technical point of view. Strategic and economic considerations are left behind. I am a scientist reconverted to engineer so please, don’t ask me for that much 🙂

Miguel Barrera

1. Data plane acceleration includes: DPDK, PCI pass-though, SRIOV ↩︎
2. IaaC is Infrastructure as a Code, and it covers any language in any format to describe your deployment, deploy and manage it. Examples are Cloud Formation, TOSCA, Terraform and others. ↩︎
3. The big three: Amazon AWS, Google Cloud and Microsoft Azure ↩︎
4. Think on Verizon, Deustche Telekom, Telefonica Tech, etc. ↩︎
5. ETSI NFV first whitepaper was launched in 2012 and others follow like ONAP in 2017. At corporate level AT&T launched ECOMP, Telefonica UNICA project, etc. ↩︎
6. An example: Amazon, Microsoft Azure and Google first news about SRIOV support are dated from 2013 to 2016 ↩︎
7. Fully managed cloud on-premises solutions like Amazon Outposts or Google Anthos could be the seed for a future far-edge enabled cloud platform technologies offered by hyperscalers ↩︎
8. These nodes collect all the traffic our mobile devices send or receive to/from the network. And remember, new radio technologies are able to offer 100Mbps-1Gbps for one single user ↩︎