chapter 27 sections

23
Chapter 27 Q and A Victor Norman CS332 Fall 2017

Upload: fathi

Post on 07-Jan-2016

24 views

Category:

Documents


1 download

DESCRIPTION

Chapter 27 sections. CS 332 Spring 2014. Link-state Routing. Q: Could you go over link-state routing (like OSPF)? - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Chapter 27 sections

Chapter 27 Q and A

Victor NormanCS332

Fall 2017

Page 2: Chapter 27 sections

Static vs. Dynamic Routes

• How are entries added to a routing/forwarding table?– via static configuration,

• by adding an interface (using ifconfig, perhaps)• by adding a route explicitly (using route add, perhaps)• by DHCP

– via a dynamic routing protocol, like OSPF or BGP.• NOTE: a routing protocol does not do packet

forwarding. It modifies the routing table dynamically.

Page 3: Chapter 27 sections

Static vs. Dynamic

Q: Does the routing algorithm care how routes were added?A: No.Q: Do hosts usually have dynamic routes added?A: No: or at least much less often than routers do.

Page 4: Chapter 27 sections

Static vs. Dynamic Routes

Q: Does Calvin use static or dynamic routes?

A: Most everything at Calvin is switched. So, most machines use DHCP to get their IP addresses and default route. This would be considered a static route. Hosts, in general, do not run any routing protocol. The (default) route to US Signal on Calvin’s main router is probably a static route.

Page 5: Chapter 27 sections

EfficiencyQ: At what size of network does static routing become less efficient than dynamic routing?

A: First, what are the adv/disadv of static vs. dynamic routing?Static:• + easy to configure; – has to be done by an administrator;

– does not adapt to changing network topology.

Dynamic:• + adapts to network changes; – have to run protocol whose messages

use bandwidth; – have to initially configure protocol.

Probably worthwhile to run routing protocol if you have multiple routers with multiple paths.

Page 6: Chapter 27 sections

Link-state vs. Distance-vector algorithms

• OSPF uses a link-state algorithm for computing best routes. – Each router builds up a picture of the network in memory – which routers

connect to each other, and which routers connect to which networks.– Each router then computes best ways for it to get to remote networks

and adds routes to its routing table.– Messages are broadcast to all routers in the network.

• RIP uses distance-vector.– Each router tells its neighbors it is directly connected to “its” networks.

Then the neighbors tell their networks they are 1 hop from those networks, etc.

– Sounds good, but there are many subtle problems when links go up and down, get moved, etc.

– Messages are sent individually between neighbor routers.

Page 7: Chapter 27 sections

OSPF and metrics

Q: What does it mean that OSPF supports metrics? How does an administrator know anything about the cost of a route?

A: A hop across a network generally costs “1” – 1 hop. OSPF advertises connectivity by how many networks a packet has to traverse. But, an administrator can manually configure certain “hops” to cost more than 1.

Page 8: Chapter 27 sections

RIP

Q: Why don’t we cover RIP?

A: RIP is old-school, man. We’ve got to move with the times.

Page 10: Chapter 27 sections

Autonomous System

Q: Can you explain what an autonomous system is?

A: Yes! It is a group of networks/routers run by one organization. It is only really necessary if you connect to the Internet with multiple service providers, because BGP uses AS numbers to coalesce routes.

Page 11: Chapter 27 sections

Why have ASes?

Q: Why do we need autonomous systems?

A: Scalability.• Recall that we don’t have routes to individual hosts (in

general), but instead to networks– Keep routing tables “small”.

• But, what if we have millions of networks?– Group networks into ASes and communicate connectivity

based on them.– ISPs have to communicate to their neighbors which networks

they host. But, this rarely changes. – What changes is how ISPs connect to each other.

Page 12: Chapter 27 sections

Security

Q: What keeps someone from injecting incorrect information via messaging to routers, and if there isn't really anything then why don't we see more of these attacks?

A: BGPv4 uses authentication and security to ensure that routers don’t accept and process messages from the bad guys.

Page 13: Chapter 27 sections
Page 14: Chapter 27 sections

Summarizing routes

Q: What does the author mean when he states that routers “summarize” routing information before passing to another autonomous system?A: Consider a router connected to 4 network: 153.106.0/24, 153.106.1/24, 153.106.2/24, 153.106.3/24. It advertises 4 networks to the world.Or, it could advertise 1 network: 153.106.0/22.

Page 15: Chapter 27 sections

Old Slides

Page 16: Chapter 27 sections

Interior Gateway Protocol

Q: What is an example of an Interior Gateway Protocol?A: OSPF and RIP are two IGPs. (IGP is just a generic term for actual protocols run within an organization. Similar for EGP.)

Page 17: Chapter 27 sections

Dynamic routes on hosts?

Q: The book says that most hosts use static routes. When would be a case when a host would use dynamic routing?

A: I can’t really think of a case. Perhaps if a host is connected to two or more networks, it might run OSPF to figure how to most efficiently get out to the Internet…

Page 18: Chapter 27 sections

Static vs. Dynamic routes

Q: Other than simplicity, are there any advantages to static routing? Why would one explicitly choose it over dynamic routing?

A: There is no advantage in terms of forwarding traffic – a route is a route is a route. To have a host get dynamic routes, it would have to run OSPF or RIP, which takes some CPU time, perhaps some configuration, etc. Much simpler to just get the route from DHCP and not mess with it.

Page 19: Chapter 27 sections

Static vs. Dynamic routes (2)

Q: A routing table is made, right, when the system boots?A: The routing table is there, but is empty, usually until the machine gets a DHCP response. Then, the interface is created and an entry goes in the routing table, and, the default route is added from the DHCP response. Q: Hosts use static and routers use dynamic?A: Yes, for the most part, I’d say. Dynamic routes get updated dynamically, without human intervention. Static routes never change.

Page 20: Chapter 27 sections

Route timeouts?

Q: When an administrator sets a route, is the route permanent or does it timeout?

A: The route is “static” – i.e., permanent. It will not be changed, updated, or replaced.

Page 21: Chapter 27 sections

OSPF and routers

Q: For OSPF to work, do all routers have to run it?A: Yes. If a router is running OSPF, it will communicate with other routers running OSPF and they will automatically figure how the network topology and how to set routes in the routing table so that there are no loops, etc. Routers that don’t participate may screw that all up.

Page 22: Chapter 27 sections

Link-state Algorithms

Q: What is link-state routing?

A: Routers figure out their neighbor routers. Then, they communicate with each other sharing who they are all connected to. Each router builds up a model of the network connectivity in memory, and then runs Dijkstra’s algorithm on it to figure out the best path to send packets to get them through the network. Whenever a change in the network happens, they all update their models and change their routes.

Page 23: Chapter 27 sections

Import routes into OSPF

Q: How can OSPF import routes?A: An edge router running BGP can learn the routes from the Internet, and then OSPF can (somehow) import those routes and advertise them throughout the internal network. (Not usually necessary, though, I would think.)