By: Tomas Fredberg, Cloud Infrastructure Reference Model developer (LFN/CNTT)
Why do we need a Reference Model?
Today, for Telco Operators, Cloud Infrastructure often means a few racks of servers on a shared Ethernet switch fabric running a single Open Stack instance for a few virtualized network functions (VNFs). While there is high interest in moving to more Cloud Native Network Functions (CNFs), they are mainly deployed on virtual machines; however due to the relatively high overhead caused by virtualization and the need for more efficient deployments, many Telecoms are looking to move to Bare Metal deployments of many of their CNFs.
The reality is that there will be a long migration period for most Telco Operators before the existing VNFs can be replaced with Cloud Native CNFs. We must realize that during this period, that for many may last up to a decade, the Operators will not want to invest in or build a separate new Cloud Infrastructure for Cloud Native applications and then manually move hardware resources between the existing and new Clouds Infrastructures as demands changes.
This calls for a shared Cloud Hardware Infrastructure with a relatively large pool of hardware resources that can simultaneously host Infrastructure as a Service with Virtual Machines and Container as a Service with K8s clusters. For many Telco Operators these different Virtualization Infrastructures will be managed by different operations groups to minimize the complexity and limit the error blast radius contained within each Virtualization Infrastructure. However, it is unpractical to statically assign and physically connect each virtual server within a relatively small pool of instances on a separate physical network dedicated to only one Virtualization Infrastructure. This could lead to a large percentage of over-specified servers in the pool and the need for hands-on reconfiguration when servers needs to be moved between Virtualization Infrastructure instances, exactly the opposite of the desired outcome for maximum efficiency.
To achieve larger and more flexible server pools hosted on multiple Virtualization Infrastructure instances that change on demand e.g. when a VNF is migrated to a CNF or when different hardware resources are needed for a particular Telco function, all the hardware resources need to be interconnected with the datacenter Ethernet switch fabric. Consequently, the shared Ethernet switch fabric should not be managed directly by any single one of the Virtualization Infrastructure instances since that would give a large fault domain and complex organizational relationships.
For these reasons, the Cloud iNfrastructure Telecom Taskforce (CNTT) part of the Linux Foundation Networking (LFN) organization, has defined a layered model supporting shared networking for its Reference Model. This model forms the basis for the two Reference Architectures where OpenStack and K8s deployments are specified.
For more background and information on the and its Reference Model I can recommend the blog “Open Cloud Infrastructure for Telecom Operators – Myth or reality?” by Walter Kozlowski and the LFN YouTube Channel Webinar “Evolution of the Cloud Infrastructure Reference Model and its Applications”.
In Part 2 of this blog, we’ll cover CNTT Reference Model Basic Concepts and Logical Architecture.