The Linux Foundation Projects
Skip to main content

By: Walter Kozlowski, Telstra, Cloud Infrastructure Reference Model Lead (LFN/CNTT)

Fact: Virtualization/containerization of network functions have moved from science experiments to mainstream technology in the telco industry.

Result: That means the Telco industry expects the technology will safely, efficiently, and relatively inexpensively support the rigorous demands of 5G, IoT, and Edge. In other words, virtualized/containerized Cloud Infrastructure is considered the basic enabler for Telco operators’ business futures.

Reality: Whoever has engaged in architecting, delivering and operating Infrastructure Cloud environments have been faced with some extremely challenging dilemmas such as:

  • How to select critical ecosystem components that are both commercially viable and technically efficient, while avoiding vendor or technology lock-in.
  • How to make the ecosystem, with its multitude of very different elements, work as a coherent system.
  • How to “build an aeroplane while flying,” i.e. , how to balance existing proven technologies with the technology evolution running at the speed of light.
  • Eventually, the final question always is, who can we turn to for the strategic advice and honest guidance that is needed to find the right architecture for our needs; how do we avoid “the blind leading the blind”.

History: The painful experience of network operators faced with such serious industry challenges led to the formation of the Cloud iNfrastructure Telco Taskforce (CNTT). Its mission from the beginning was to help operators, technology suppliers and innovators alike in this rough journey. The goal seems simple, yet it is important: to show the way by building a generic, flexible and yet concrete model that spawns a select number of reference architectures that can then be implemented and validated through a defined industry compliance process.

Achievements: CNTT has created an initial Reference Model (RM), defined infrastructure and workload profiles, capabilities and measurements and mapped them together. When building the first version of the RM, we had in mind a VM-based (using OpenStack) architecture. We proceeded to building upon this model and published an OpenStack based Reference Architecture (RA-1) which was implemented through (RI-1) and will soon be possible to test through the RC-1 compliance mechanisms.

The work has already started to help the industry move forward. There are already examples where using the RM defined profiles and relationships significantly sped up the creation of RFXs (request for something) by providing a standard way of asking questions and answering them. It also helped to use a standard and thought-through way of assigning network workloads to an infrastructure profile. The resulting RA-1 was used as a guide for building a concrete architecture for some operators. RI-1 (and RC-1) are quickly becoming a recognised playground for trying some solutions and possibly, eventually for badging them.

Challenges: With the speed of the technology evolution, tomorrow is already upon us. We have built a VM-based reference specification, while the industry was jumping into the cloud native, containerised world. In recognition of this fact, CNTT started a process of making the Reference Model more generic to be able to support RA-2, a Kubernetes based Reference Architecture and respective RI/RC-2. Most of the operators are already working with an always evolving coexistence of many virtualization (IaaS) and containerization (CaaS) environments sharing the same hardware (compute, storage, network) environments. This is means there is a need for the infrastructure to be managed by different administrative domains in a very complex multi-tenant ecosystem. The extreme performance requirements typical for 5G, IoT or distributed network applications (Edge), are driving the need for more details on how these architectures will work from a practical perspective. So far, the CNTT specifications were missing answers to questions on how to integrate SDN into the Cloud Infrastructure, how to automate build and operations of network fabric, how to deal with the various acceleration and offloading technologies, how to evolve towards the future programmable fabric, and how to adopt and adapt cloud native techniques, so successful in the IT world, into the realities of the telco networks.

CNTT Response: Through analysis of real life situations, CNTT has already identified several gaps in its Reference Model, and for that matter, in all existing industry models (e.g. around missing open Hardware Infrastructure Management). CNTT is working with some other LFN projects, such as ODIM (Open Distributed Infrastructure Management), and other industry organisations to address these issues. The Reference Model is currently being extended to cater to container-based networking co-existence and migration scenarios, and is looking into incorporating acceleration and programmable fabric technologies. The CNTT Baraque Release (September 2020) Reference Model includes an early version of the extended model as presented in the following realization diagram.

The biggest change is the introduction of a new component – Hardware Infrastructure Manager— to fill the gap as there are no current, common sets of HW Infrastructure Managers on the market or in any open source community that can handle multiple equipment vendor environments in a way that is not tightly connected to a specific Virtualization/Containerization Infrastructure. Note also how the Model supports virtual infrastructure implementations managed from different administrative domains. The Model also envisages the use of multiple SDN controllers working on different levels, and finally allows for direct management access (e.g. from the Orchestration layer) into the hardware infrastructure layer if this is done via the open and standardised HW Infrastructure Manager. This has the potential to allow for the introduction of new technologies such as P4-based programmable fabrics.

Future: While no one knows exactly what the future will bring, CNTT needs to keep evolving the Reference Architectures and corresponding Reference Implementations in line with the Reference Model evolution. The success of this project largely depends on the community involvement in this work. Your passion and expertise will be an invaluable contribution!

To learn more about this topic, please view the following webinar: “Evolution of the Cloud Infrastructure Reference Model and its Applications”. To learn more about participating in CNTT, please view the wiki here and email