# Computing in the Network: The Core-Edge Continuum in 6G Network

> chapter 10 of the book "Shaping Future of 6G network"

## Intro
* 6G帶來的影響
* resource utilization
* performance
* new services with applications in IoT, manufacturing and autonomous system

* 跟5G相比，整個網路的基本架構並沒有多大的變動，但是網路有逐漸的
* clouderization(? 這不知道是啥)
* softwarization(SDN)
* Virtualization

* 有些新穎的概念被提出

* End2End encryption
* QUIC(Quick UDP Internet Connection)
* White box as network nodes

* 而隨著近期研究主軸的改變，目前data centers(DCs) 還有edge networks相比Internet本身更被關注著，使用AI、ML等data driven 的技巧來解決network上的優化問題已經時有所聞
* 讓網路可以自主的決定要用怎麼樣的algo、怎麼樣的架構來確保性能

⇒ Computing in the network, 同時computing 以及networking的議題出現各式各樣的網路中
* 以往在設計網路的時候比較少出現computing task 跟networking整合的情況，在多數情況都是:
> A relative dumb networking enabling forwarding between Intelligent entities
>
在這種設計方法之下，networking並沒有甚麼運算上的要求，他只要專心負責追求速度以及安全性即可。
而如今我們卻要求network node需要包含computing的功能，顯然這會造成我們想要**Keep network node simple and connsequently very fast**的需求有些背道而馳。
幸虧現在有了硬體技術上的進步，同時我們也增加了**virtualization**以及**data processing**的技巧，因此只要我們能夠預先地進行一些local processing，或許就得以試著解決這個從根源上矛盾的問題。

~~'While 5G defined the concept...~~ 太長了 後面就跳過owo

## Milestones of  Programmable Networks

事實上Computing in network並不是一個新議題，畢竟一開始所謂的network node的任務就是:
> Computers used to implement connectivity protocols and determining routes by executing routing algos

雖然為了要確保**reliability**, **speed**, **security**，通常我們會minimize generic computations, 這25年來人們還是不斷的努力，希望可以在這個web-centric的時代提升**programmability**

### Active network
> Adapt the network to dynamic requirements by adding executables on demand in nodes traversed by packets.
針對目前到來的packets, 使用不一樣的executables來應對需求

**Executables 來自哪裡?**
有兩種可能的方法
* 在到達的packet當中包含有對應的檔案
* 儲存在router當中

可惜的是這樣的想法並沒能很好的落地，因為要解讀packet當中的東西，這本身就會有security上的疑慮。

### Information-centric Networking
> ~~好長 看不懂 估計可能用不到 就先跳過吧:P~~

### Compute-first Networking
系統可以知道現在剩餘的資源量來決定要啟用router當中的哪一個funciton
* 算立

### Software-defined Networking
>The move to cloud-based computing and the rise of DCs has been significant in the evolution of networks.

傳統上的routing是由硬體上決定的，而SDN將整個流程分隔為兩個平台，在軟體的平台上決定routing的行為，而實際傳輸資料的部分則是交由硬體完成。
因為我們這個將操作分隔開來的操作，我們對網路的**programability**就會尚稱，因為我們要新增service的時候不再需要艱難地去改動我們的硬體，只需要在軟體上完成修改就好。

## The Computarization of Network
大家發現，硬體上越簡單，越有可能在軟體上給予網路可操控的能力跟彈性，也因此漸漸地就出現了**The softwarization of Networking**
而如今因為網路扮演的已經不再只是單純的傳遞資料的角色，而是同時還需要負責一些processing的步驟，這就被稱為**The computerization of networking**
往下我們會分別針對這些新增加的角色做點描述

### A New End to End Paradigm
> Most of the requirements for the network services today include mobility, scalability, security, and availability, often in real-time or in a timely manner, anywhere and over a wide variety of end devices.

過去使用的架構( cloud and client-server architecs) 在解決自動化、決策這種需要快速做出反映的問題的時候，會有複雜度上的問題
:::info
cloud and client-server 的意思就是server 在cloud上面，當client有需要的時候，就把東西都發給server，讓她計算完再回應
:::
在edge resource 做簡單基礎的前處理
在cloud端做advanced computing
如此一來可以減輕cloud端的壓力，針對request可以快速地給出回應，就算是有delay sensitive的application也可以放心使用

### Computing in the Network: basic concepts

#### definition
the execution of native host applications within the net node

#### Focused part
* over the network layer: the cloud and the SDN virtual node
* under the network layer: the mobile and wireless infrastructure
* At the edge of network: applications and devices to support them

#### In/On-network
* In-network(directly on the nodes)
* On-network(in associated peripherals such as storage node)

### Related Impacts
為了要做到**packet processing** and **associate function beyond forwarding**, 勢必會有一些需求或新技術會隨之而生
* Need for resource discovery:
如果不能動態的察覺現有的可用資源有哪些，那就沒有辦法很有效率的調整自己的行為，因此這一項技術是必須的

* Power saving for Eco-conscious Networking
資料的傳輸是一件非常消耗能源的事情，有研究指出相比於運算，在傳輸過程當中所消耗的能源大約是運算的2倍或以上
> 70% ~ 90% of the power consumed in a data system is consumed to transport that information
Thus, distributing the computing inside the network especially on existing devices is energy saving

雖然分散式的運算可能會進一步有延遲上的疑慮，但限如今運算設備的性能越來越強，這一點疑慮也逐漸可以被消除

* Transport is still needed
adding computing to transport layer will:
* Alleviate congestion with local execution
* Reduce network loads with local caching, data reduction, compressed sensing, and pre-rendering

* How about security
The suitability of privacy and security needs to be evaluated for the robustness to attacks

## Core-edge Continuum

### Features
* Disaggregation(這個是啥 看不懂)

* Multitenancy
可以輕鬆地將必要的資料共享，同時能確保裝置之間的隔離性
shared data easily while preserve isolation between devices
* Data reduction
Taking advantages of computing and storage resource to deal with large amount of data generated between network and edges

* (Near-)real-time
With functionality constantly adapting in response to mobility, workload, and application requirements. It can provide low latency

### Structure

* Common Data Layer
* The New Programmable Data Plane
* Novel Architectures Using Computing in the Network

## Making it Real: Use Cases

* XR
* IoT
* Autonomous decision system
* Computing in the Data Center
* Data and Flow Aggregation
* Key-value Storage and In-network Caching
* Consensus
* Next-generation IoT and Intelligence Everywhere
* The Internet of Intelligent Things
* Industrial Automation: From Factories to Farms
* Computing Support for Networked　Multimedia
* Video Analytics
* Extended Reality and Multimedia

* Melding AI and Computing for Measuring and Managing the Network
* AI/ML for Network Management
* Telemetry

* Network Coding

### Computing in the Data Center

# Questions
1. What is Data center(DC)
2. 傳統router使用硬體做routing是怎麼做到
3. Edge network? (感覺跟network edge不是一樣的東西)
4. What is cloud and client server structure
5. Disaggregation

Outline
=======

1. Intro
: list the difference of 6G and 5G
(three main differences on the aspect of Software)
3. Problem formulation
: These differences are contradict to the core concept of network(The simpler the better)
5. Solutions from past to nowadays
: How people implementing Computing in network, pros and cons
7. Conclusion of Solutions
: Some of the Solutions works well nowadays and make some good impacts
9. Intro to the novel concept offer by author
: the structure of core-edge continuum, actually I'm not willing to talk about this part since it's too vague inside the article.
But since it's the topic of this article, maybe We can't avoid mentioning it:<
11. Case in Reality
12. Conclusion