# 3 July 2020 - Daily Report

## Summary
----
### Expected Outcome :
* Find out more about AI/ML utilization in O-RAN
* Find out how O-RAN utilize Acumos platform
___
### Outcome :
* Get a better view about Acumos AI platform
* Understand the place where AI/ML is utilized
* Understand the basic of RIC

### Further plan:
* Study more about RIC

___
### Timeline
**08.55** - Signed In
**09.00** - Writing study plan
**09.10** - Starting study
**11:40** - Discussing with Kevin about study plan
**12:49** - Praying and having lunch
**13:57** - Start to study about RIC
**15:56** - Praying
**16:15** - Start to study and write the report

---
## Study Notes

### Acumos
>Reference :
>* https://www.youtube.com/watch?v=zvBpsbE-0fQ
>* [My acumos notes](https://hackmd.io/@akmalns/rJlEiDwx8)

The video reference above give a brief explanation of how acumos works. From the video i get more information that has not covered in my previous notes.

#### **Why do acumos exist?**

**Picture Source** : Screenshoted from video reference

In the past few years, Artificial Intelligence usage and development increased rapidly alongside with the growing computational power we can achive. Many company and groups that start to utilize Artificial Intelligence or Machine Learning model to automate or assist their works. They also seek for the minimal cost for the deployment.

So the Acumos AI platform was born to make a model interchangeable from each other hence reducing time and cost needed to make a brand new model. Using acumos, we could also on-board our model and train it with data that is collected from data scientist/data miner. The model will be stored as a microservices in the Acumos platform. This microservices is interoperable to make a more complex system of AI/ML Model. Acumos is a platform for AI/ML modeller, data miner and user who use AI/ML.

#### **Platform User Flow**
**Picture Source** : Screenshoted from video reference

The picture above is showing the user flow in acumos platform. The flow begin from the AI/ML modeller who make AI/ML using different tools like tensorflow, scikit-learn, etc. The AI/ML model will be on-boarded to the platform as a microservices. It will be stored in microservice repository. Each of this microservices can be chain together by the analyst to make a more complex system of AI/ML. After chained together, the model will be validated by operator before enter the catalog and can be seen by future user.
#### **Microservices**
**Picture Source** : Screenshoted from video reference

We can see from the picture that the complete model of video content analysis consists of many model. They are tailored together so form a full functional model. Each of the part is a microservices. The microservices is model that is made by ML modeller.

### Discussion with kevin summary
The discussion take place in skype. We discuss about study plan mainly detailing what we need to study. From the discussion, we can conclude that the first thing we need to study is
1. O-RAN (Link to [my notes](https://hackmd.io/@akmalns/HJHrJh5AL))
2. RIC (I find the brief description but not the details. Link to [my notes](https://hackmd.io/@akmalns/HJHrJh5AL))
3. Acumos
4. Kubernetes

We also need to submit daily notes link to Kevin/Derni so that they can check it.
About the on-boarding error we encountered during model on-boarding to Acumos, there are several problems that lead to the error. The problems are:
1. We need to deploy acumos first before on-boarding.
2. Acumos used to be deployed under a docker but the current All-In-One(AIO) deploy is not supporting docker anymore.
3. Acumos need to be deployed under kubernetes although a small error is still existing

We will try to deploy acumos together after finishing current study plan.

### RIC
>Reference:
> *http://www.techplayon.com/5g-ric-ran-intelligent-controller/

#### What is RIC?
RIC or RAN Intelligence Controler is a suite of Software Apps that enables SDN (Software Defined Network) to works in open RAN networks. RIC can be operated in any RAN operation and optimization procedure like radio connection management, mobility management, QoS management, edge services, interference management, etc. RIC can utilize AI?ML models to support the best effective network opertion.

#### Types of RIC

There are two types of RIC that are Non-Real Time RIC and Near-Real Time RIC.
The picture above shows where the Non-RT and Near-RT located in O-RAN.

#### Non-Real-Time RIC (Non-RT RIC)

The Non-RT RIC functions are service and policy management, RAN analytics and model-training for the Near-RT RIC. It also control policy, rules, and other network parameter in RAN.

#### Near-Real-Time RIC (Near-RT RIC)

The Near-RT RIC provides new functions leveraging embedded intelligence such as QoS management connectivity management and seamless handover control. It also a platform that allows for flexible on-boarding of third-party control-applications. It also provides initiate configuration commands to CU/DU. Near-RT RIC control of RAN element in near-real-time control on a fine-grained network operation such as radio resource management, load-balancing, ID allocations, handover decisions, etc.

#### RIC in O-RAN
>Reference:
>* O-RAN use cases and deployment, Feb 2020. It can be accessed [here](https://static1.squarespace.com/static/5ad774cce74940d7115044b0/t/5e95a0a306c6ab2d1cbca4d3/1586864301196/O-RAN+Use+Cases+and+Deployment+Scenarios+Whitepaper+February+2020.pdf)
>* O-RAN.WG2.Use-Case-Requirements-v02.00, I get it from Ferlinda. Can be accessed in O-RAN website.

Continuing from the [previous notes](https://hackmd.io/@akmalns/HJHrJh5AL). From the O-RAN.WG2.Use-Case-Requirements-v02.00, The Non-RT RIC can be involved in some use cases such as Traffic steering use case, QoE use case, QoS based resource optimization, etc.

___

## Conclusion
* Acumos is a platform where many model from different modeller meets so that it can be trained, combined and deployed on many platform.
* RIC is the part in O-RAN that utilize AI/ML model to control many things in the RAN

###### tags: `Daily Report`