Transcript ppt

ECEN5553 Telecom Systems
Week #4
Readings:
Read [5a] "Whatever Happened to the IPv4 Address Crisis?
thru
[5c] "How Can the Internet Have Too Many Routes
and Not Enough Addresses"
Read [6] "The Cognitive Net is Coming"
Exam #1: Lecture 14, 16 September (Live)
No later than 23 September (Remote DL)
4 page test. Work 3.
1-2 pages will be off Fall 2015 Exam #1
Outline: Lecture 22, 5 October (Live)
No later than 12 October (Remote DL)
Outlines
Received
due 5 October (local)
12 October (remote)
5%
IEEE Ethernet

802.3 10 Mbps (1983)
 Coax
→ Twisted Pair
 Shared → Switched
 Half Duplex → Full Duplex
802.3u 100 Mbps Fast Ethernet (1995)
 802.3z 1 Gbps Ethernet (1998)
 802.3ae 10 Gbps Ethernet (2002)
 802.3ba 40 & 100 Gbps Ethernet (2010)
 802.3bs 400 Gbps Ethernet (2017?)

Performance Issues

Throughput
 Usable

BW
Efficiency
 Percent

of time packets are moved
Shared Ethernet Efficiency
 100%
under Low Load
 ≈ 1/(1+5*NPD) under Heavy Load
Low Speed Network?
This configuration...
PC
Server
Hub
PC
PC
10BaseT & Shared Hub
... is as good as this one...
PC
Server
Switched
Hub
PC
PC
10BaseT & Switched Hub


...IF traffic mostly going to/from same machine
Switched Hub better if diverse traffic flow
This configuration is even better.
PC
10 Mbps
1 Gbps
Server
Switch
100 Mbps
PC

10 Mbps
PC
Server on a higher speed line.
Shared Ethernet Efficiency
Gbps has higher NPD
1.0
0.5
0.0
.01
.10
.1
1.0
NPD
1
10.0
10
100
High Speed Network?
This configuration has
horrible throughput.
PC
Server
Shared
Hub
PC
PC
1 or 10 Gbps & Shared Hub
Under heavy load, too much time
spent recovering from collisions.
Ethernet (Shared) Hub
Operates at OSI Level 1
 ‘Electric Cable’
Traffic arriving at an input is immediately
copied to all other ports on a bit-by bit basis.
 Used on LAN's. Pretty much obsolete.
 Repeater = single input & single output hub

 Not
used much on Ethernet any more
 Generally now only used on WAN long haul
 May
be different protocol than Ethernet
Black Box Performance...
From
Node A
Node B
Node C
To
Node A
OSI Level 1
LAN Hub
Node B
Node C
Two packets simultaneously show up at input...
Black Box Performance...
From
Node A
Node B
Node C
To
Node A
OSI Level 1
LAN Hub
Node B
Node C
... will overwrite each other, i.e. garbage out.
a.k.a. Shared Hub
Black Box Performance...
OSI Level
1-2 (Switch)
or 1-3 (Router)
Black Box Performance...
OSI Level
1-2 (Switch)
or 1-3 (Router)
Two packets simultaneously show up at input
& need to exit on same output...
Black Box Performance...
OSI Level
1-2 (Switch)
or 1-3 (Router)
... one will be transmitted (when allowed by MAC),
the other momentarily buffered...
Black Box Performance...
OSI Level
1-2 (Switch)
or 1-3 (Router)
... and then transmitted.
10BaseT & Shared Hub
53 m
PC
8m
PC
Hub
PC
26 m
17 m
One big collision domain.
PC
10BaseT & Half Duplex Switch
53 m
PC
8m
PC
Switch
PC
26 m
17 m
PC
Four smaller collision domains.
Example
If Box 1 & 2 are Level 1 Hubs
One Big Collision Domain
15 Nodes share 10 Mbps
Each node gets average of 10/15 Mbps
World
10BaseT
7
Hub 1
Users
Hub 2
7
Users
Right Side to World gets
7/15th of available BW,
on average.
Example
If Box 1 & 2 are Level 2 Switches
Each node shares 10 Mbps with Switch
Right Hand Side is on one 10 Mbps line.
World
10BaseT
7
Users
Sw 1
Right hand side sees
increased delays.
Can be alleviated with
100 Mbps Box 1 ↔ Box 2 link.
Sw 2
7
Users
Right Side to World gets
1/8th of available BW.
_____
(Was 7/15th)
Switched Hubs or Bridge
On Power Up know nothing
 When a packet arrives at an input port...

 Look-Up
Table consulted
 Source MAC address not in table?
 Table
Updated: MAC address & Port matched
 Destination
 Packet
broadcast to all outputs (a.k.a. flooding)
 Destination
 Packet

MAC address not in table?
MAC address in table?
shipped to proper output
Look-up Table update is dependent on
packet arrivals
Router
Operates at OSI Layers 1, 2, & 3
 Capable of making complex routing decisions

 ‘peers
into’ packets and examines Layer 3 address
Very useful on Large Networks with multiple
end-to-end paths
 Routers frequently exchange connectivity info
with neighboring Routers

 Routing
Algorithms used to update
Routing (Look-Up) Tables
 Tables updated independently of traffic
Bridging versus Routing

Ethernet Bridge, Switch, or Switched Hub
 Uses
Layer 2 MAC Address
 Unknown Destination? Flooded
 Look-up Table updates are packet dependent

Router
 Uses
Layer 3 Internet Protocol Address
 Unknown Destination? Default location
 Look-up Tables updated independently of traffic
Small Network? Doesn't matter
 Big Network? Floods not a good idea.

Ethernet Switch
Uses MAC Source Address to populate Look-Up Table
Uses MAC Destination Address & Table for I/O Decision
Bytes: 7
1
6
6
2
Pre
SFD
Destination
Address
Source
Address
46-1500
Data + Padding
Len
4
CRC
Router
Populates Look-Up table independently of traffic.
Uses Destination IP Address & Table for I/O Decision
Bytes: 7
1
6
6
2
MAC
Destination
Address
MAC
Source
Address
Does look at the MAC Address! Is this for me?
20
20
6-1460
4
IPv4
TCP
Data + Padding
CRC
Data Network Addressing

IP Address
 Global
Information Source
 Global Information Sink
 Stays unchanged end-to-end
 Exception: Network Address Translation

Ethernet Address
 Local
Transmitter (MAC Source)
 Local Receiver (MAC Receiver)
 Crossing a Router (Ethernet Boundary)?
 MAC Header and Trailer swapped out
 Crossing a Switch?
 MAC Header and Trailer unchanged
Multiplexing
Sharing a chunk of Bandwidth
by splitting it into channels
 Channel can carry one conversation


FDM, TDM, & StatMux
FDM
Different channels use some of
the frequency all of the time.
frequency
1
2
3
4
5
TDM
Different channels use all of
the frequency some of the time.
Fixed, predictable times.
frequency
1
2
3
1
StatMux
Different channels use all of
the frequency some of the time,
at random, as needed.
frequency
1
3
1
2
StatMux vs. TDM & FDM
uses bandwidth more efficiently for
bursty traffic
 requires more overhead
 has more variable deliveries
 requires more complex hardware

Switching: In what manner will
a user get to use a channel?
For the duration of the conversation?
Circuit Switching
 For a tiny, variable length, portion of the
conversation?
Packet Switching
 Circuit vs. Packet Switching
Circuit has less end-to-end delay
Circuit is less complex
Packet is more efficient for Bursty Traffic

MULTIPLEXING
SWITCHING
StatMux
Circuit
Packet
TDM
FDM
X
X
Any Switching & Multiplexing combo possible.
Two marked are among most common today.
LAN/MAN History: FDDI
(Fiber Distributed Data Interface)
Developed in ’87 – ‘88
 Covered OSI Layers 1 & 2
 1st 100 Mbps Line Speed
 Token Ring MAC

 Guaranteed
Bandwidth
Had Priorities.
 Originally Dual Counter-Rotating Rings

Designed for Metropolitan Area
Counter Rotating Fiber Rings
Outside Active.
Inside Hot Standby.
Designed for Metropolitan Area
Counter Rotating Fiber Rings
1
Line
Break...
Nodes 1 & 4 wrap.
One big ring.
4
FDDI Status
 Never
succeeded as a LAN
 NIC's
 Saw
use mostly as a corporate backbone
 OSU
 Was
too expensive
backbone from 1989 - 1993 ish
fairly common at Internet Exchanges
 Used
to pass traffic
from ISP A to ISP B
 Now too slow
1993 OSU Stillwater Network
(15)
(21)
The Internet
VAST collection of interconnected
networks
 Key Building Block:
Routers running IP (Layer 3)
 Router link speeds range up to 200 Gbps
 Hierarchical Alpha-Numeric Names
[email protected]

AT&T 1997 Internet Backbone
UUNET 1998 Internet Backbone
AT&T 2009 Internet Backbone
Source: http://www.business.att.com/content/productbrochures/MIS_15906.pdf
Washington D.C. Area - 2000
OSU 2009 Internet Connectivity
Traceroute to WWW.CISCO.COM





3 Internal OSU-Stillwater routers
4 OneNet routers (all in OKC? Tulsa?)
3 Qwest routers
dal-edge-18.inet.qwest.net
Akamai Technologies (Hosting Service)
(11:51 am, 9Sept15, rtt = 13 msec, 10 routers)
(12:32 pm, 9Sept16, rtt = 13 msec, 10 routers)
Traceroute to WWW.TULSA.COM



3 Internal OSU-Stillwater routers
3 OneNet routers (Tulsa)
5 Cogent Communications routers
 te0-0-1-7.rcr21.tul01.atlas.cogentco.com
 be2706.ccr21.mci01.atlas.cogentco.com
(Kansas City?)
 be3035.ccr21.den01.atlas.cogentco.com
 be3037.ccr21.slc01.atlas.cogentco.com

4 Unified Layer routers (Hosting Service)



prv-211-1-0-1.unifiedlayer.com
End server (198.57.177.235) in Provo, Utah area?
(12:43 pm, 9Sept16, rtt = 45 msec, 15 routers)
ISP Routes Sometimes Roundabout
Launched 13 September 2014, 2 miles from OSU campus
 1 Scheets' home router
 4 AT&T routers
 adsl-70-233-159-254.dsl.okcyok.sbcglobal.net
 ggr3.dlstx.ip.att.net
 4 Cogent Communications routers
 Be2032.ccr22.dfw01.atlat.cogentco.com
 te0-0-2-1.rcr12.okc01.atlas.cogentco.com
 3 ONENET routers
 OKC?
 3 Oklahoma State routers
 (12:30 pm, 11Sept14, rtt = 84 msec, 15 routers)
Fall 2007 Weird TraceRoute Seen by Student
Tulsa to OSU Stillwater
 Tracert launched from Tulsa, hit
Atlanta
Washington, D.C.
Illinois
Kansas City
Tulsa
Oklahoma City
OSU Stillwater
Internet Service Provider Backbone
Router
Switched Network, full duplex trunks.
Access lines attach to corporate routers &
routers of other ISP's.