接上第二点

NOTE

(2)active networks => Programmability in networks(1990s)

Sturcture:
What are active networks?
motivation for active networks
technologies behind active networks
How do active networks related to SDN?
The legacy of active networks

What are active networks?

1.Networks where switches perform custom computations on packets.

2.Examples(and motivation)

-1.Trace program running at each router. (i.e. traceroute)

-2.Middleboxes, eg.firewalls, proxies(代理), application services. <=> custom computations on traffic, Middleboxes <=> modern-day active networks

Origins of active networks.

1.DARPA research community(1994-1995), i.e.discussing

2.Identified problems with today's networks

-1.Difficulty of integrating new technology. => it's very difficult to develop technologies on today's network.

-2.Poor performance due to redundant(冗余) operations at several protocol layers(protocol stacks)

-3.Difficulty accommodating(容纳,适应) new services.

Motivation for active networks

1.Accelerating innovations(the same as SDN)

-1.Internet innovations rely on consensus(一致)

-2.Take ten years from prototype to deployment.(standardlization => procurement(获得,i.e.生成新的设备) => deployment)

2.Active nodes allow routers to download new services into the infrastructure. => User-driven innovation

Idea: Message carry Procedures and Data



(pics from wikipedia)

1.Active routers coexist(共存) with legacy(旧的) routers.

2.Each programmable switch can perform additional processing.

User "Pulls" and Technologies "Push"

1.User Pulls(demand)

-1.Proliferation(增加) of firewalls, proxies, transcoders, etc.

-2.Goal: Replace ad hoc(which means "for this" in Latin) approaches.

2.Technologies Push(enablers) => provide services to meet the users' demands

-1.Safe excecution of mobile code, Java applets.

-2.OS Support: Scout(real-time communications), Exo_kernel(safe access to low-level resources), SPIN(trustworthy code generation) => provide better environment and services => nice support

Two different approaches to active networks

1.Capsules("integrated") => every packets carried a program => node excecute

-1.Every message is a program. Active nodes evaluate(评估) content in packets.

-2.Code dispatch(迅速处理) to excecution env.

数据报即程序,积极节点执行处理并转发。

2.Programmable Switches("discrete(分离的)") => nowadays

-1.Custom processing functions run on routers.

-2.Packets are routed through programmable nodes.

-3.Program depends on the packets headers.

节点根据数据包首部和提前写入节点的程序进行处理。like p4 nowadays.

2017/2/23

Coursera SDN M1.2.1 SDN History: Programmable Networks 1的更多相关文章

  1. Coursera SDN M1.2.1 SDN History: Programmable Networks 2

    NOTE Capsules Example 1.Type Field: excute forwarding routine => carry program 2.Previous address ...

  2. The Road to SDN: An Intellectual History of Programmable Networks

    文章名称:The Road to SDN: An Intellectual History of Programmable Networks 文章来源:Feamster N , Rexford J , ...

  3. 【论文】The Road to SDN: An Intellectual History of Programmable Networks

    目录 ABSTRACT: 1 Introduction: 2 The Road to SDN: 2.1 Active Networking Technology push and use pull I ...

  4. Coursera SDN M1.1 SDN History: Central Control

    source Structure 1.讨论SDN的时间线,从1980s至今. 2.认识到SDN背后的原则和idea. 3.识别SDN起源的架构主题. NOTE Four Chapter in SDN ...

  5. SDN 编程语言 p4(SDN programming language P4)

    行业趋势,SND是未来. P4 是未来. SDN is inevitably, and P4 is inevitably. P4 = Programming Protocol-Independent ...

  6. SDN期末作业-通过SDN的应用实现负载均衡

    负载均衡程序 1.程序链接:https://github.com/424baopu/software/tree/master/LoadBalance 2.场景 topo: 场景描述: 服务器host ...

  7. 【原】Coursera—Andrew Ng机器学习—课程笔记 Lecture 9_Neural Networks learning

    神经网络的学习(Neural Networks: Learning) 9.1 代价函数 Cost Function 参考视频: 9 - 1 - Cost Function (7 min).mkv 假设 ...

  8. 【原】Coursera—Andrew Ng机器学习—Week 5 习题—Neural Networks learning

    课上习题 [1]代价函数 [2]代价函数计算 [3] [4]矩阵的向量化 [5]梯度校验 Answer:(1.013 -0.993) / 0.02 = 3.001 [6]梯度校验 Answer:学习的 ...

  9. 【原】Coursera—Andrew Ng机器学习—课程笔记 Lecture 8_Neural Networks Representation 神经网络的表述

    神经网络是一种受大脑工作原理启发的模式. 它在许多应用中广泛使用:当您的手机解释并理解您的语音命令时,很可能是神经网络正在帮助理解您的语音; 当您兑现支票时,自动读取数字的机器也使用神经网络. 8.1 ...

随机推荐

  1. Python学习之k-近邻实例

    海伦收集约会数据巳经有了一段时间,她把这些数据存放在文本文件datingTestSet.txt中,每个样本数据占据一行,总共有 1000 行.海伦的样本主要包含以下 3 种特征: 1. 每年获得的飞行 ...

  2. Codeforces Round #375 (Div. 2)E. One-Way Reform

    题目链接:传送门 题目大意:一副无向图,要求你给边定向(变为有向图),使出度等于入度的点最多,输出有多少 个点,并且输出定向后的边(前为起点,后为终点) 题目思路:欧拉路 我们这样考虑,先考虑无向图的 ...

  3. 关于layer.photos即照片显示的问题。

    在layer组件中,照片显示是不常用,今天做了一些不伤了. 在这里写出来,以备后用. 其中注意几个问题, 1.格式问题. 2.路径问题. 不同的layer有不同的格式,查看layerAPI中发现的格式 ...

  4. POJ 3461 Oulipo[附KMP算法详细流程讲解]

      E - Oulipo Time Limit:1000MS     Memory Limit:65536KB     64bit IO Format:%I64d & %I64u Submit ...

  5. hibernate在Oracle中插入数据,默认字段被设置为null的问题解决

    参考内容: http://blog.sina.cn/dpool/blog/s/blog_90629d5301014a5w.html 在数据库中一个字段的默认值为1,但是在插入数据后,本来该字段为空,值 ...

  6. Swift - 获取当前系统时间

    // 获取当前系统时间 let date = NSDate() let timeFormatter = NSDateFormatter() timeFormatter.dateFormat = &qu ...

  7. 优雅的go语言--入门篇

    1.特点 1.静态类型,编译型的开源语言 2.脚本华的语法,支持多种编程范式(函数式&面向对象) 3.原生,给力的并发编程的支持 2.优势 1.脚本化的语法 2.静态类型+编译型,程序运行速度 ...

  8. 树形DP求各点能走到的最远距离

    hdu2196 Computer Time Limit: 1000/1000 MS (Java/Others)    Memory Limit: 32768/32768 K (Java/Others) ...

  9. [Android] 配置安卓模拟器,使得dex文件不被优化成odex

    最近做一个模块,需要将apk里面加载的dex文件dump出来,所以需要配置让dalvik不要把dex文件优化成odex. 1. 配置build.prop 主要是通过修改文件/system/build. ...

  10. HDU 5618 Jam's problem again(三维偏序,CDQ分治,树状数组,线段树)

    Jam's problem again Time Limit: 5000/2500 MS (Java/Others)    Memory Limit: 65536/65536 K (Java/Othe ...