1.2.1 Access network
DSL 连接 DSLAM，类似拨号上网，占用电话线路。
客户端：splitter seoarates the data and telephone signals arriving to the home and forwards the data signal to the DSL modem.
telco side: in CO, DSLAM分离诗句，并传输到i纽特网。
asymmetric
Cable Internet access 需要cable modems. 头部CMTS于DSLAM类似，将模拟信号从cable modem转为数字信号。将HFC转成两个频道 downstream and upstream。
LAN： ethernet switch -> institutional router - to institution's ISP.
WLAN: wireless LAN(local access network) based on IEEE 802.11
1.3
1.3.1 Packet switching
router and layer switches
store-and-forward transmission: 传输前要收到所有的包。同时排了几个包的话，在接收的同时可以发送已经接收完成的包 d_end to end = NL/R
queue delays and packet loss: 每个packet swith has multiple links attached to it. 每个link, the packet swith has an output buffer(queue), which 存储了要传送进link的包。如果包要传输，但当前链接在处理其他包，就要排队。因此在store ... delays 以外，还要suffer output buffer queueing delays. 看拥堵时间了。如果等待位满了还传包，会发生packet loss，会drop掉arriving packet（新的数据）或queue packet
假定A,B用自己带宽超高的电脑向C传数据，数据会成为queue of packet waiting for output link，最终通过的速度会慢。和corcuit switching不一样，不保证速度，但能多快多块
Forwarding tables and routing protocals: 等等再看这部分
1.3.2 Circuit Switching
在传输前，分配一条路径，就像传统的电话网络一样。有constant transmission rate in the network's links for the duration of the connection.
Multiplexing in cuicuit switched network - FDM和TDM两种。对FDM而言，是根据频率不同，会分到不同的link。电话是4KHZ bandwidth. FM 88-108MHZ. 对TDM而言，时间被分为固定的间断，每个frame被分为固定数量的slots。当被分配一个connection accross a link时，这个网络会向每个frame的固定time slot发送信息。
支持packet switching的人会觉得circuit switching是一种浪费，在silent periods。并且觉得end to end circuit难以搭建，保证end to end transmission capacity是需要复杂的signaling software来配合这种path的。
PS vs CS
packet switching is not suitable for real-time services, because of its variable and unpredictable end to end delays.
反对意见：it offer better sharing of transmission capacity. 并且 it is simpler more efficient and less costly to implement than circuit switching .
说到这里为什么PS更有效率呢？举个栗子，若用户共享1M带宽，假设每个用户都间断性使用网络。或当一个用户在稳定使用100K带宽并且会间断性的不活跃。或一个用户只有10%的时间活跃。CS是极度没有效率的，若CS TDM传输，10 time slot， 不论是大于10个用户导致带宽无法分配，还是小于10个用户导致带宽浪费，都是不好的。
1.3.3 A network of networks
access ISP 使用DSL， cable, FTTH, WIFI以及cellular等技术连接网络。不局限于telco或cable company，还可以是公司或者学校的局域网。ISP需要interconnected 即network of networks。
所有A ISP之间可以互相传送数据
涉及到几个结构：
1） network structure 1 : 将所有A ISP连接到a single global transit ISP上。这个大ISP起路由的作用。customer ->provider
2) network structure 2: A ISP连接multipleglobal transit ISPs。 但全球ISP互相要连接。
tier-1 ISPs 《- regional ISP
3）network structure 3: multi-tier hierarchy
4) network structure 4 :
POP（point of presence） (除了A ISP没有，其他都有) is a group of one or more routers in the provider's network where cusomer ISPs can connect into the provider ISP. (路由表)
muilti-home 连接多个ISP
peer 同级互连
IXP（Internet Exchange point）：a meeting point where multiple ISPs can peer together.
5) network structure 5: adding content provider networks
整理一下： Access ISP 连接 regional ISP / IXP / Tier 1 ISP / Content provider , 请想象阶乘的曲线，全能连接。
1.4 performance metrics: delay, throughput, packet loss
architect 3_ lec
no delay constraint = packet loos probability close to zero
delay(node) = delay(processing) +delay(queue) + delay(transmission) + delay(propagation)
delay(processing) is defined by hardware
delay(transmission) = packet size / transmission rate
delay(propagation) = distance / propagation speed
congestion 是大于capacity的部分
arrive rate = 1/mean interarrival time
以上延迟都是DETERMINISTIC！！！
delay(queue) = N(t) = multiplexer switch network T Seconds RANDOM!!!!!
Packet loss: 当buffer waiting area 满了时， packet arriving to full buffer is lost.
Throughput: bits / second, 计算一定时间窗口内的均值
Min(Rs, Rc) 如果有两个上限，肯定取最小的那个啊
3个Rs -> R -> 3个Rc Min(Rs, R/3, Rc)
Efficiency: Data 成功送到目的地 除以 Overall resource used.(包括overhead, idle time/ frenquency, lost packets)
1.5 layering: motivation, ISO stack, internet stack, encapsulation
motivation:provide functionalities/services to the upper layers ; some information exchange ; not optimal
OSI stack:
Internet layer: Content(7) - > Application(5-7) -> Transport(4) -> Network(3) -> Physical & Link"LAN-link"(1-2)
OSI layer
Layer 7: application : implemented by hosts (http, pop, smtp, ftp, telnet)
Layer 4: transport: end to end control(TCP/UDP)
Layer 3: network: end to end packet forwarding
Layer 2: link: link scale services
Layer 1: physical (bits <=> signal)
encapsulation:
3.6.1 congestion and scenarios
Application Architecture:
client-server: