D
Dredgie
This is a contributed article by Stéphane Téral, Chief Analyst at LightCounting and founder of TÉRAL RESEARCH
Although the programmability of the novel Open RAN (O-RAN) architecture provides great flexibility and agility, its predefined telemetry and control features remain the last of the bottlenecks to tackle. The essence of Open RAN lies in the virtualization of RAN functions, open interfaces between decomposed RAN elements, and full access to telemetric data to foster much needed innovation by developers. But the fact is, in the existing 800 LTE networks - including the 265 of them recently upgraded to 5G - detailed telemetry is widespread across network nodes and components, except in the RAN! The rise of Open RAN is changing the game.
In cooperation with Capgemini and Intel, Microsoft developed a programmable RAN platform framework that embeds virtualized RAN, analytics, and automation. Illustrated in Figure 1, Microsoft’s proposed architecture enables the introduction of flexible, dynamically loadable service models and allows developers to upload their own or 3rd party custom code that safely and directly implements a service model1 onto a live RAN platform.
Microsoft adds the concept of the “D” - the dynamic service model, to the o-ran architecture
As Figure 1 suggests, the proposed architecture is based on the existing O-RAN reference framework, including its RAN Intelligent Controller (RIC), but with a tweak: A “D”, as in dynamic service model. The beauty of this D functionality is that it can be implemented by the application developer and dynamically deployed at run time with no impact on RAN operations. As a result, communications service providers (CSPs) and their trusted 3rd party developers can write their own telemetry, control, and codelets, and deploy them at various points in the live RAN software stack without any operational interferences. This is a very important new feature because the codelets are executed inline on the critical path of the live RAN. This gives them access to the whole internal raw RAN data structure, affording the ability to collect statistics and make real-time inference and control decisions.
The analytics and automation framework is carefully designed with reliability, safety, and security in mind. The code used to define dynamic service models is statically verified by using state-of-the-art verification technologies. It relies on the same tools used to run user space Extended Berkeley Packet Filter (eBPF) codelets in the Linux kernel on millions of mission-critical servers around the globe. The code is automatically preempted if it runs longer than its predefined execution budget. It is also extremely fast, typically incurring less than 1 percent overhead on the existing RAN operations.
Implementing what can be considered Real-Time RIC functionality, Microsoft’s controller features a fast, sub-10 ms, control loop (Figure 1) to address the need to support new applications enabled by the dynamic service model. And of course, this analytics and automation framework meets all the high reliability and top safety and security criteria.
The “D” is set to boost RAN energy savings, which remain a top priority
It’s no breaking news that RAN energy savings has been an issue since the beginning of mobile networking. When the first 4G LTE rollouts started in 2010, self-organizing network (SON) energy features were introduced. Now in our fifth year of 5G, RAN energy savings continue to be a top priority for all CSPs in the world. The added analytical flexibility that the D provides can significantly improve existing RAN energy-saving techniques, including SON algorithms, which are not typically easy and fast to deploy. This is because it allows fine-grained access to many metrics across different layers of the RAN stack. This means that for energy prediction, a developer can just collect the data they need instead of sorting and analyzing giant sets of data sent to the xApp of the RIC system.
With its initial prototype built on a Capgemini 5G RAN and Intel’s FlexRAN reference software, Microsoft’s initial prototype delivered 30% energy savings overall, including in busy hours. Given the additional hundreds of millions of dollars large CSPs in the world have seen on their energy bill, this 30% savings is more than welcome.
Bottom line: after energy savings, only the sky is the limit!
This is just the start of a very long journey: the migration of 30-year-old traditional RAN design to a virtualized, disaggregated, one with open interfaces and a novel controller (RIC) augmented by Microsoft’s dynamic service model. Only the radios cannot move to software, the rest can and will, opening the door to a plethora of innovative applications.
Although many use cases such as energy saving, coverage and capacity optimization, and handover are not necessarily new, having being implemented with SON, it’s worth mentioning that not every CSP chooses to adopt the SON option. Case in point: approximately two thirds of commercial LTE networks operate without SON. However, we are now living in a 5G world - a more complex architecture that generates a tsunami of telemetries and key parameter indicators - and humans are no longer able to keep up. And with the arrival of RAN virtualization, RAN disaggregation and open interfaces, the integration of multiple RAN components and functions further increases the level of complexity, which further complicates anomaly detection. Consequently, there is no way back, RAN programmability is about to rule the world.
To learn more about Microsoft’s approach to programmable RAN, check out this Azure white paper.
Stéphane Téral Stelyana Baleva
Chief Analyst at LightCounting Principal Analyst at LightCounting
Founder of TÉRAL RESEARCH Executive Director of Research at TÉRAL RESEARCH
stephane@teralresearch.com stelyana@teralresearch.com
---
Endnote.
1. A RAN service model is a way of describing how the radio access network operates and interacts with other network elements and services. It can define the architecture, interfaces, algorithms, security, and performance of the RAN, as well as the deployment and management options. Today’s service models are inherently static, in nature, with committees within the O-RAN Alliance tasked with reviewing and approving any changes to them. ~SD
Continue reading...
Although the programmability of the novel Open RAN (O-RAN) architecture provides great flexibility and agility, its predefined telemetry and control features remain the last of the bottlenecks to tackle. The essence of Open RAN lies in the virtualization of RAN functions, open interfaces between decomposed RAN elements, and full access to telemetric data to foster much needed innovation by developers. But the fact is, in the existing 800 LTE networks - including the 265 of them recently upgraded to 5G - detailed telemetry is widespread across network nodes and components, except in the RAN! The rise of Open RAN is changing the game.
In cooperation with Capgemini and Intel, Microsoft developed a programmable RAN platform framework that embeds virtualized RAN, analytics, and automation. Illustrated in Figure 1, Microsoft’s proposed architecture enables the introduction of flexible, dynamically loadable service models and allows developers to upload their own or 3rd party custom code that safely and directly implements a service model1 onto a live RAN platform.
Figure 1: Microsoft’s proposed dynamic service model architecture (Source: Microsoft)
Microsoft adds the concept of the “D” - the dynamic service model, to the o-ran architecture
As Figure 1 suggests, the proposed architecture is based on the existing O-RAN reference framework, including its RAN Intelligent Controller (RIC), but with a tweak: A “D”, as in dynamic service model. The beauty of this D functionality is that it can be implemented by the application developer and dynamically deployed at run time with no impact on RAN operations. As a result, communications service providers (CSPs) and their trusted 3rd party developers can write their own telemetry, control, and codelets, and deploy them at various points in the live RAN software stack without any operational interferences. This is a very important new feature because the codelets are executed inline on the critical path of the live RAN. This gives them access to the whole internal raw RAN data structure, affording the ability to collect statistics and make real-time inference and control decisions.
The analytics and automation framework is carefully designed with reliability, safety, and security in mind. The code used to define dynamic service models is statically verified by using state-of-the-art verification technologies. It relies on the same tools used to run user space Extended Berkeley Packet Filter (eBPF) codelets in the Linux kernel on millions of mission-critical servers around the globe. The code is automatically preempted if it runs longer than its predefined execution budget. It is also extremely fast, typically incurring less than 1 percent overhead on the existing RAN operations.
Implementing what can be considered Real-Time RIC functionality, Microsoft’s controller features a fast, sub-10 ms, control loop (Figure 1) to address the need to support new applications enabled by the dynamic service model. And of course, this analytics and automation framework meets all the high reliability and top safety and security criteria.
The “D” is set to boost RAN energy savings, which remain a top priority
It’s no breaking news that RAN energy savings has been an issue since the beginning of mobile networking. When the first 4G LTE rollouts started in 2010, self-organizing network (SON) energy features were introduced. Now in our fifth year of 5G, RAN energy savings continue to be a top priority for all CSPs in the world. The added analytical flexibility that the D provides can significantly improve existing RAN energy-saving techniques, including SON algorithms, which are not typically easy and fast to deploy. This is because it allows fine-grained access to many metrics across different layers of the RAN stack. This means that for energy prediction, a developer can just collect the data they need instead of sorting and analyzing giant sets of data sent to the xApp of the RIC system.
With its initial prototype built on a Capgemini 5G RAN and Intel’s FlexRAN reference software, Microsoft’s initial prototype delivered 30% energy savings overall, including in busy hours. Given the additional hundreds of millions of dollars large CSPs in the world have seen on their energy bill, this 30% savings is more than welcome.
Bottom line: after energy savings, only the sky is the limit!
This is just the start of a very long journey: the migration of 30-year-old traditional RAN design to a virtualized, disaggregated, one with open interfaces and a novel controller (RIC) augmented by Microsoft’s dynamic service model. Only the radios cannot move to software, the rest can and will, opening the door to a plethora of innovative applications.
Although many use cases such as energy saving, coverage and capacity optimization, and handover are not necessarily new, having being implemented with SON, it’s worth mentioning that not every CSP chooses to adopt the SON option. Case in point: approximately two thirds of commercial LTE networks operate without SON. However, we are now living in a 5G world - a more complex architecture that generates a tsunami of telemetries and key parameter indicators - and humans are no longer able to keep up. And with the arrival of RAN virtualization, RAN disaggregation and open interfaces, the integration of multiple RAN components and functions further increases the level of complexity, which further complicates anomaly detection. Consequently, there is no way back, RAN programmability is about to rule the world.
To learn more about Microsoft’s approach to programmable RAN, check out this Azure white paper.
Stéphane Téral Stelyana Baleva
Chief Analyst at LightCounting Principal Analyst at LightCounting
Founder of TÉRAL RESEARCH Executive Director of Research at TÉRAL RESEARCH
stephane@teralresearch.com stelyana@teralresearch.com
---
Endnote.
1. A RAN service model is a way of describing how the radio access network operates and interacts with other network elements and services. It can define the architecture, interfaces, algorithms, security, and performance of the RAN, as well as the deployment and management options. Today’s service models are inherently static, in nature, with committees within the O-RAN Alliance tasked with reviewing and approving any changes to them. ~SD
Continue reading...