# Toward Next Generation Open Radio Access Networks--What O-RAN Can and Cannot Do!
###### tags: `5G Reading`
Date : 2022-10-07

## Metadata
[paper link](https://arxiv.org/pdf/2111.13754.pdf)
Abdalla, A. S., Upadhyaya, P. S., Shah, V. K., & Marojevic, V. (2022). Toward Next Generation Open Radio Access Networks--What O-RAN Can and Cannot Do!. IEEE Network.

## Take away
**What is O-RAN**
Provides virtualization, intelligence, and flexibility while defining open interfaces for network innovation in 5G.

**What O-RAN can do:**
A. RAN Disaggregation, Open Interfaces, and Multi-Vendor Support
B. Support for Different Timescales
C. AI Integration and xApps/rApps

**What O-RAN can notdo:**
A. End-to-End Security
B. Deterministic Latency
C. PHY Layer RT Control
D. Testing of AI-Driven O-RAN Functionalities

## Summary
O-RAN defines a disruptive architecture for realizing next-generation wireless networks. It specifies open interfaces and logical elements that support intelligence, disaggregation, softwarization, virtualization, and collaboration. The study conducted **a survey among researchers, developers, and practitioners to explore their interests in O-RAN and their needs for 6G R&D**.
As a result of the survey, they identify critical limitations of the current O-RAN specifications in terms of **security, latency, real-time control, and testing of AI-based RAN control applications**.
They also outline the technologies and R&D opportunities to overcome these limitations and extend the O-RAN architectural capabilities.

## Note
- Participants in this study
In the survey done in this study, the data was collected from 2020 to 2021. The participants are from 65 institutions; 50 located in North America, 12 in Europe, and 3 in Asia. There were 95 total participants: 21 full professors, 13 associate professors, 14 assistant professors, 9 postdoctoral researchers, 10 experienced PhD candidates, and 28 experts from industry, government research laboratories, and research institutes.

- O-RAN Overall architecture

- What O-RAN can: A. RAN Disaggregation, Open Interfaces, and Multi-Vendor Support
In 3GPP spec., they define the functional split of the network architecture, however, the vandors still use their proprietary implementations and
interfaces. O-RAN breaks the situation by splitting them into O-CU, O-DU and O-RU, that enables cellular operators to employ unique strengths of each vendor for a delicate application or service and facilitates RAN function sharing.

- What O-RAN can: B. Support for Different Timescales
O-RAN uniquely introduces two timescales control loops which are the non-RT RIC(O1, A1) and near-RT RIC(E2).

- What O-RAN can: C. AI Integration and xApps/rApps
The non-RT RIC is a controller of the SMO that provides policy management, AI management, and data enrichment services to the underlying RAN nodes. Furthermore, besides collecting and training AI models for xApps in the near-RT RIC, the non-RT RIC can itself host AI-driven rApps, such as AI-based orchestration of network slices and data-driven policy management, to improve the automation and management of RAN resources and xApps.
---
- What O-RAN can not: A. End-to-End Security
The O-RAN architecture enables the dynamic disaggregation of functionalities, the introduction of new functions, protocols, components, and interfaces, however, it also exposes the architecture to more security risks.

- What O-RAN can not: B. Deterministic Latency
Open interfaces and multi-vendor support spur innovation.
This, however, makes it more difficult to control and optimize the data and control plane latency. The general concept and latency model of the O-RAN architecture is based on the eCPRI reference model for delay management.

- What O-RAN can not: C. PHY Layer RT Control
While the near-RT RIC could perform rudimentary control of PHY layer functions, such as reducing the beam search space, the timescale at which it operates (10 ms – 1 s) makes it difficult to control many of the PHY layer processes. Promising solutions to the above problem include running the zApps at near-constant time complexity combined with subject-matter expert validation and developing a new class of highly accurate and lightweight AI algorithms, such as echo state networks, which are less data intensive. The integration of RAN hardware acceleration within O-RAN is expected to augment the performance of x86 based server platforms that implement O-CUs and O-DUs.
- What O-RAN can not: D. Testing of AI-Driven O-RAN Functionalities
The unpredictable behavior of the AI algorithms, specifically when considering a closed-loop AI controller, may lead to unstable system configurations and performance losses. Integrating services into the AI workflow for such testing and validation and into next-generation RAN processes for delivering high-fidelity live representations of all functions that are running on O-RAN deployments, where one O-CU, O-DU, and O-RU configuration may differ from another within the same operator and across operators and services.