CS 3640: Introduction to Networks and Their Applications Fall 2018, - - PowerPoint PPT Presentation

1

CS 3640: Introduction to Networks and Their Applications

Fall 2018, Lecture 8: The Link Layer III – Addressing Protocols (animations from Christo Wilson @ NEU) Instructor: Rishab Nithyanand Teaching Assistant: Md. Kowsar Hossain

2

You should…

• Be checking Piazza regularly for announcements.
• Be close to making a submission for assignment 1.
• Know and understand:
• The three Internet design principles.
• Encapsulation.
• The components of the Internet.
• Circuit- vs. packet- switched networks.
• Components of end-to-end delay.
• Functions of the link layer.
• How error detection works at the link layer.

Recap: Medium access control protocols

• What is the function of a MAC protocol? What are the

general strategies for designing one?

• Regulate access to the shared medium.
• Strategies: Medium partitioning, token-based access, best-effort need-

based access.

• What are the problems with partitioning and token-based access?
• Why does slotted ALOHA have better performance than

standard ALOHA?

• Introducing “slots” means that the “sensitive” transmission period is cut

in half. This basically doubles the achievable throughput.

The link layer: The CSMA/CA MAC protocol (1990s)

• Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA).
• Currently used in wireless transmission protocols (WiFi 802.11).
• In wireless networks:
• Carrier sensing while transmitting is not feasible, so collisions can only be detected

at the receiver. Why?

• Detecting collisions is harder because accurately sensing the medium is tough –

wireless signals may not carry to far away hosts.

• Connectivity is not transitive: If A can reach B and B can reach C, it doesn’t mean that A

can reach C.

• Instead, wireless networks just try to avoid collisions instead of detecting and
• retransmitting. They use the CSMA/CA protocol.

The link layer: The CSMA/CA protocol

• B wants to transmit to C.
• Sense the carrier. Is another node transmitting?
• Yes. So do an exponential back-off before trying again.

A B C D

The link layer: The CSMA/CA protocol

• B wants to transmit to C.
• Sense the carrier. Is another node transmitting?
• No. Send frames. If I don’t get an ACK in reasonable time, start again.

A B C D

The link layer: The CSMA/CA MAC protocol

• Step 1:
• Sense the carrier. Is another node transmitting?
• Step 2:
• If the carrier is not busy: Send the frames.
• If the carrier is busy: Do an exponential back-off and go to step 1.
• Step 3:
• Wait for an acknowledgement. If it doesn’t arrive after “timeout” seconds,

go to step 1 and try again.

• Problem: Collisions can still occur and utilization can be low!
• Why? Hidden nodes and exposed terminals.

The link layer: The CSMA/CA MAC protocol – Hidden terminals

• Step 1: Sense the carrier. Is another node transmitting?
• Hidden nodes can increase false-negatives.

A B C

Collision! A cannot hear C C cannot hear A

The link layer: The CSMA/CA MAC protocol – Hidden terminals

• Discuss: Would asking for permission from the receiver help?

A B C

Collision! A cannot hear C C cannot hear A

The link layer: The CSMA/CA MAC protocol (with RTS/CTS)

A B C

The link layer: The CSMA/CA MAC protocol (with RTS/CTS)

• Step 1:
• Sense the carrier. Is another node transmitting?
• Step 2:
• If the carrier is not busy: Tell the receiver you have something to send via a

“Request To Send” (RTS) message. Wait for “timeout” seconds for a “Clear To Send” (CTS) message from the receiver.

• If CTS arrives: send the frames.
• If CTS doesn’t arrive: do an exponential back-off and go to step 1.
• If the carrier is busy: Do an exponential back-off and go to step 1.
• Step 3:
• Wait for an acknowledgement. If it doesn’t arrive after “timeout” seconds, go to step

1 and try again.

• Solves the hidden terminal problem, doesn’t help with exposed

terminals.

The link layer: The CSMA/CA MAC protocol – Exposed terminals

Carrier sense detects a busy channel

No collision

A B C D

No collision

• Step 1: Sense the carrier. Is another node transmitting?
• Exposed terminals can increase false-positives.

The link layer: The CSMA/CA MAC protocol – Exposed terminals

A B C D

• Discuss: Why doesn’t RTS/CTS messaging work here?
• RTS/CTS helps reduce false-negatives (thinking that the medium is free when it

isn’t) but doesn’t prevent false-positives (thinking the medium is busy when it isn’t) because we never actually reach the RTS stage when we have a false-positive.

The link layer: The CSMA/CA MAC protocol – Exposed terminals

• Discuss: Why not skip sensing and just start with an RTS

message if it solves the exposed terminal problem?

• Overhead!
• RTS/CTS packets don’t have any useful data in them. As your network gets

faster, you send more RTS/CTS packets per second because you can handle more transmissions per second.

• RTS/CTS overhead on a 1Mbps WiFi connection: 4%
• RTS/CTS overhead on a 11Mbps WiFi connection: 25%
• Sending RTS/CTS packets when you don’t need to will only make this

worse.

• Lesson: Everything is a trade-off!
• Researchers and engineers decided that the overhead from this approach was

too high to be a feasible solution to the exposed terminal problem.

Recap: Medium access control protocols

• What is the general idea behind CSMA protocols?
• Sense the medium before transmitting.
• Why are wireless media more challenging?
• Collisions can only be detected at the receiver and carrier sensing is not

always accurate.

• Why do we need a CSMA/CD and CSMA/CA protocol?
• Collisions can still occur due to transmission delays, CSMA/CD detects

these and improves throughput by stopping transmission as soon as it is detected.

• CSMA/CA tries to avoid collisions entirely by requesting permission

from the receiver before transmission.

16

This week in class

1.

Recap: Error detection in the link layer

2. 3.

Link layer principles: Medium access Link layer principles: Addressing

Addressing in the link layer

• When we studied MAC protocols, we assumed that packets

went on every wire (i.e., they were broadcasted).

• This is why collisions occurred, but there was no need for any logic in the

link-layer hardware.

• Originally, Ethernet was a broadcast-only medium and that’s how things
• worked. Today, with new types of hardware we have new opportunities for

efficiency-gain.

Terminator Repeater

Link layer and network interface technologies: Hubs

Hub

C B

Send Packet B → C

• Discuss: What are the pros and cons of using Hubs for

your LAN?

• Pros: Dirt cheap hardware with no logic.
• Cons: Poor efficiency and scalability. Good enough for small LANs.

Link layer and network interface technologies: Bridges and switches

• Discuss: What could we do to improve efficiency and

scalability?

• Stop using broadcast! Only send packets to their intended recipients.
• This is what bridges and switches do. They contain logic to figure out

which wire to put packets on.

A C B

Send Packet B → C

Bridge

Link layer and network interface technologies: Bridges and hubs in the real world

• In the real world, we use a combination of old ethernet

broadcast nets, hubs, and bridges.

• Bridges are usually used to create a “bridge” between two

small LANs. Hub Hub

Link layer and network interface technologies: Bridges and hubs in the real world

• Discuss: Why aren’t bridges used within all LANs?
• This is the real world. You cannot force deployment – even when it is

better (Windows XP in most federal agency computers).

• Bridges are more expensive and complex than hubs.

Hub Hub

Link layer and network interface technologies: Bridges vs. Hubs (the internals)

• Why are bridges more expensive and complicated?
• They have memory for packets to be queued.
• The have internal logic to forward packets to the correct wires.

Switch Fabric

Inputs Outputs Bridge Makes routing decisions Hub Memory buffer

How do switches/bridges know what packet goes on which wire?

• Each interface has a “MAC address”.
• This is a 48-bit string that is allocated by the device manufacturer.
• MAC addresses are unique to a device and usually cannot be changed.
• Regulated by the IEEE.
• Every link layer frame needs to have a source and destination MAC

address to ensure forwarding is done correctly.

How do switches/bridges know what packet goes on which wire?

• Each interface has a “MAC address”.
• Each bridge/switch maintains a “forwarding table”.
• Basically a collection of records of which MAC addresses are

accessible through which ports. Records expire after a “timeout”.

00:00:00:00:00:DD 1 3 minutes MAC address port age 00:00:00:00:00:AA 1 1 minute 00:00:00:00:00:BB 2 7 minutes 00:00:00:00:00:CC 3 2 seconds

How do switches/bridges know what packet goes on which wire?

• How is the forwarding table built?
• Look at the source of each arriving frame.
• Which MAC address sent the frame?
• Which port did it reach the bridge on?

Hub

00:00:00:00:00:AA 00:00:00:00:00:BB Port 1 Port 2 00:00:00:00:00:BB 2 0 minutes MAC Address Port Age 00:00:00:00:00:AA 1 0 minutes

Delete old entries after a timeout

How do switches/bridges know what packet goes on which wire?

• How is the forwarding table used?
• If the frame has a destination MAC address that is in the forwarding

table, send the frame out the corresponding port.

• If the frame has a destination MAC address that is not in the

forwarding table broadcast the frame through all ports (except the incoming one).

Port 1 Port 3 Port 2 Port 4

Forwarding tables in action

• FF

EE DD CC BB AA Port 1 Port 2 Port 1 Port 2

Hub Hub Hub

AA 1 AA 1 CC 2 CC 1 EE 2 EE 2

Bridge 1 Bridge 2

How loops can break forwarding protocols

• <Src=AA, Dest=DD>
• This continues to infinity.
• Discuss: How do we stop this?
• Remove loops from the topology.
• We use an algorithm to build and

maintain a spanning tree for link layer addressing.

AA Port 1

Hub

Port 1

Hub

Port 2 Port 2 AA 1 AA 1 BB CC DD AA 2 AA 2 AA 1 AA 1

What is a spanning tree?

• A subset of edges in a graph that:
• Spans all nodes.
• Does not create any cycles.

1 4 2 5 6 3 7 1 4 2 5 6 3 7 5 1 4 2 6 3 7

How do spanning trees help?

• If bridges in our topology can collectively organize their forwarding

tables to make the topology seem like a spanning tree, then all loops are removed.

1 4 2 5 6 3 7 1 4 2 5 6 3 7 5 1 4 2 6 3 7

How do nodes construct a spanning tree from any network topology?

• All bridges randomly elect a single bridge as the “root” of the

spanning tree.

• Example (shortest straw): All bridges pick a random number. The bridge with

the smallest random number is the root.

• Each bridge finds the shortest path to this root.
• Problem: Need to know who the root is!
• Solution: Broadcast your best knowledge to all neighbors. Update your
• knowledge. Repeat until steady state is reached.
• The union of all these paths is a spanning tree.

How do nodes construct a spanning tree from any network topology?

0: 0/0 12: 12/0 3: 3/0 27: 27/0 41: 41/0 9: 9/0 68: 68/0 27: 0/1 12: 0/1 41: 3/1 68: 9/1 41: 0/2 3: 0/2 68: 3/2 9: 3/2 68: 0/3 9: 0/3

Bridges vs. Switches

• Both make it possible to increase LAN capacity via the same

approaches.

• Automatically learn and maintain forwarding tables.
• A switch is a special case of a bridge.
• Bridge: Each port can be connected to either another bridge, hub, or broadcast

net.

• Switch: Each port can only be connected to a single device (an end-host or

another switch).

• You don’t need MAC protocols in switches! Why?

How does the link layer frame travel from source to destination within a LAN?

• We know how frames are forwarded within the LAN.
• Switches and bridges. Both rely on knowing the destination MAC address.
• But how does the sender know the address of the receiver?
• The Address Resolution Protocol (ARP)
• Each end-host maintains an ARP table.
• This is a collection of <IP address, MAC address, TTL> tuples.
• When a packet from the network layer arrives, the link layer looks at the

destination IP address and fetches the corresponding record from the ARP table.

• Discuss: What if there is no entry in the ARP table?
• Broadcast an ARP request asking for a response from the end-host owning the destination IP

address.

• ARP response has MAC address.
• More in Assignment 2!

Discussion

• Could the whole Internet be one big switching domain? What

would this look like? What issues would appear?

• Constant broadcasting to locate unknown hosts (billions of these!) would be a

disaster!

• Reaching a steady state with the spanning tree would be very improbable.
• Each switch would need to know every MAC address on the Internet! Think of

the memory that would required!

• We use IP addressing and network-layer routing to avoid these

problems.

• Topic for the next few weeks!9999999999p999