Vendor View

Dynamic network functions placement, an essential capability for successful 5G?

This Vendor View is sponsored by Amdocs


Recent research by AvidThink (leading independent research and analysis firm for infrastructure technologies) explores the challenges of homing and network functions (NF) placement, recommending a three-phase approach that will lead service providers on a step-wise journey to autonomous NF placement.

Today’s NF placement orchestration capabilities typically address NF resource allocation by simply duplicating resources across the core and multiple edge locations to deal with peak hour service usage. While network functions resource duplication might be appropriate as a short-term workaround, this will no longer be a cost-effective solution as 5G networks are deployed at scale.

NF placement must be optimized if 5G return on investment is to be maximized, which is where next generation homing and dynamic placement capabilities come into play. Such capabilities will enable service providers to optimize resource allocation, while avoiding the expenses of resource duplication, as well as upfront investments in network topologies designed for peak hour usage. At the same time, these capabilities mean that service providers can fully deliver on performance and user experience expectations.

AvidThink recommends that for service providers rolling out a disaggregated 5G network, NF placement optimization will significantly improve their business case. As these players build out the network for 5G core and populate the RAN with 5G new radios, now is the opportune time to update the foundations of their orchestration system and upgrade to one that can enable the next generation of services that 5G will bring.


A world of possibilities is opening up with the new, enhanced capabilities of 5G around capacity throughput, latency and volume of connected devices. New opportunities range from smart sensor monitoring and fixed wireless access, all the way to autonomous vehicles, smart cities, remote surgery and much more. Indeed, it is evident that the true potential of 5G is based on its ability to enable, deliver, measure and charge for differentiated services. 5G network slicing in particular allows services to be configured to the specific needs of disparate customers, applications and industry verticals.

Building 5G networks and realizing the full potential of those networks to drive innovative revenue streams comes at considerable cost. To ensure an effective return on investment, service providers must not only identify compelling revenue opportunities for 5G but also optimize their networks to reduce costs. Here, NF placement, the cornerstone of 5G disaggregated architecture, has a critical part to play by enabling the optimal resource allocation that can accommodate a wide range of differentiated services.


5G networks, unlike 3G/4G, can be tailored or “sliced” to meet the needs of specific use cases, services, applications and customers. By leveraging 5G network slicing, service providers can split the physical network into multiple, virtual networks. This new ability, which enables support of multiple new use cases, as well as the ability to monetize differentiated services by committing to a guaranteed quality of service (QoS) over a network slice, makes NF placement optimization more important than ever before.

For example, to provide and guarantee the low latency that real-time applications like interactive mobile gaming or industry 4.0 require, there is a need to allocate more application processing power or “brainpower” closer to the end user to shorten the delay between the time the end-user device triggers a request to the time it gets an answer back from an application in the core network. In existing NF placement orchestration systems, the application controller is moved from the service provider’s mobile core to the edge locations. But since the end user in a mobile network is nomadic/mobile, the application in all edge data center locations would need to be duplicated in order to provide guaranteed quality of experience. Such duplication of resources is both inefficient and expensive.


A more efficient and cost-effective solution is an adaptive NF placement system that optimizes NF functions placement and allocates network resources in real time according to actual demand. Such a solution will also take multiple parameters into account, such as available resources, geographic network topology, user locations, application/service/slice requirements and targeted service-level-agreements (SLA). To achieve this level of efficiency, NF placement needs sufficient “business integration context” or intent-based intelligence to translate the business input that defines a differentiated service into the underlying parameters for the NF placement orchestrator.

While terms like “optimization”, “automation” and “intent-based” are all too familiar and hold great promise, where should service providers begin? In their Amdocs-sponsored research paper, AvidThink outlines the three phases of dynamic placement and suggests pragmatic actions that service providers can take to move towards dynamic NF placement, while addressing the challenges and implications.


What information does the NF placement orchestrator need in the move towards dynamic NF placement? How will the orchestrator know that it needs to change the placement of a network function? What input does it need to make a decision?

This current stage of reliance on workarounds is referred to in the research paper as ‘static’ or ‘pre- instantiated’ NF placement. In today’s homing and NF placement systems, NF orchestrators rely on workarounds, or resource duplication and hardware optimization techniques, to inform NF placement orchestration. Examples include: The Linux Foundation’s (HPA), which allows the orchestrator to utilize information from the underlying NFV-infrastructure; virtual infrastructure managers (VIM) like OpenStack; and containers managers like Kubernetes.

This is phase one, where NF resource planning is based on peak-hour load. During this phase, network functions are placed once only, and the orchestrator uses load balancers and basic policy rules to split the load on the network between available resources in real time. Meanwhile, service providers use their own resources at the edge data center or in some cases, utilize public cloud resources to avoid the necessity of significant upfront investment in data centers.

The compelling event to move to the second phase, that of “adaptive” and optimized NF placement, is the introduction of 5G disaggregated architecture and the dynamic nature of scalable virtual and cloud network functions (VNFs and CNFs).

While NF placement optimization over network resources in a dynamically changing environment like a mobile network is a major challenge, it has a significant role to play in the successful implementation of 5G networks. Here network slicing acts as a critical enabler to the operation and automation of a network with differentiated services. Yet this also introduces a significant level of complexity due to the requirement for guaranteed isolation and agreed levels of QoS and latency for each slice. In this phase of NF placement, artificial intelligence and machine learning (AI/ML) assist in identifying slice utilization patterns and alert the NF placement orchestrator of thresholds crossed before actual QoS degradation occurs.

The adaptive nature described in this second phase could actually be implemented with ‘semi-automatic’ NF placement, where the NF placement orchestrator chooses from two or three pre-instantiated locations, in real time and based on predefined policy rules and real-time AI/ML. This predefined and monitored environment enables a higher level of placement autonomy based on a pre-measured set of network capabilities and policy rules. This can be compared with the well-known 5 stages on the journey to autonomous driving. There are a number of interim “conditional automation” stages between “no driving automation” and “fully autonomous driving” where the car, or in our case, the NF placement orchestrator, can perform most of the driving (or placement) tasks, but with human override or controlled environment still required.

At the stage of adaptive NF placement, the orchestrator must be able to support closed-loop operations, using feedback from telemetry and KPIs to support placement decisions. Where necessary, network functions can then be re-allocated to maintain the guaranteed QoS and committed SLAs for each service or slice. In the autonomous driving analogy, this closed loop capability is an essential stage in the “transformation of responsibility” journey, where the NF placement orchestrator becomes increasingly autonomous in a controlled environment.

And so to the third and final phase, which is fully autonomous NF placement. In this phase, the NF placement orchestrator handles the entire lifecycle of the service or slice with its associated NF resources, using AI/ML as needed to fine-tune dynamic placement. Here, the ability to dynamically orchestrate and re-allocate NFs is key to further optimizing and improving real-time placement over private and public clouds and even over multi-operator networks as required.


5G networks require a complete shift in the way networks are managed, as well as in the way services are deployed and rolled out. Service providers can start laying the foundations today. Bridging the gap between the network layer and the business enablement stack is essential as a basis for the intent-based NF placement intelligence that is necessary to translate the business input that defines a differentiated service into the underlying parameters for the orchestrator. With this as a starting point, service providers will be in a strong position to manage the challenges of 5G deployment and successfully monetize 5G.

For more information, download the AvidThink report which is sponsored by Amdocs.

Visit for a range of resources on 5G network slicing.

Leave a comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.