Chapter 4 Network Layer Flashcards
(136 cards)
1
Q
The two types of packet switched networks?
A
Datagram
Virtual Circuit
2
Q
Datagram
A
- Packets that carry full destination addresses
- Packets of same flow can follow different paths
- Can arrive out of order
3
Q
Virtual Circuit
A
- Packets carry a Virtual Circuit ID.
- Packets of same flow contrained to the SAME path
4
Q
What type of routing using the datagram apporach?
A
End-to-end routing and point-to-point
5
Q
Routing
A
Process of deciding routes to use and populate routing tables.
6
Q
Where is the virtual circuit model still used?
A
Within ISP Networks, in MPLS Multi Protocol Label Switching
7
Q
Forwarding
A
The process of a router using its local table to determine where an incoming packet is coming from and where it should go to?
8
Q
Who builds the router tables
A
Each router should build its own table using distributed algorithms
9
Q
Fate Sharing
A
The motivating principle for routers to build their own tables. If the router fails its responsible for its own routing info
10
Q
How does Software Defined Networking change routing
A
A central controller decides routing rules and pushes them to routers
11
Q
Objectives in routing?
A
Achieve good performance
- Short Paths
- Spreading traffic among network links to make good use of resources, load balancing
12
Q
Policy-based routing
A
Routing decisions based on commerical or organizational rules, not performance
13
Q
Adaptivity in networks
A
- Adaptive to changes in topology
- Adaptive to traffic load changes
14
Q
Slosh Effect
A
Links may oscillate between being too heavily loaded and being too lightly loaded
15
Q
The top objectives in routing
A
Stability
Correcting
Fairness
Fault tolerance
Low overhead
16
Q
Stability in routing
A
Don’t want oscillations, should converge quickly after topology changes
17
Q
Intra-AS Routing
A
- Distributed: each router builds its own table
- Each link has an assigned weight
- Routing aims to find shortest paths
- Autnomous Systems
18
Q
How are link costs determined?
A
- On-line adaptive manner
- Fixed length (often 1)
- Inverse of capacity (1/capacity)
- Off-line tuned (Traffic Engineering)
19
Q
Two Main Types of Routing Algorithms for traffic engineering
A
- Distance Vector
- Link State
20
Q
Distance Vector Routing used by?
A
Routers
- - Routing Information Protocol
- EIGRP (Cisco Enchanced IGRP)
- Original ARPANET Routing
21
Q
Bellman-Ford Equation
A
Dx(y) = min( Dv(y) + c(x,v) )
min(go to y + x to v )
22
Q
“Count-to-infinity” Problem
A
Routers increment path cost repeatedly after failure due to stale infor loeading to slow convergence and routing loops
23
Q
Distance vector routing reacts quickly to ____ but may suffer from very slow converage after ____
A
- Link Restoration
- Link Failure
24
Q
Hold-down strategy
A
Ignore better-looking routes after detecting a longer path, giving time for bad news to spread
25
Poisoned Reverse
if router x routes to y via z, it advertuses that the path to y is infinite, preventing 2 node loops
- Works for simple linear topology
26
What protocols use Link state routing
Appoarch used in IS-IS (Intermediate system to intermediate system) and OSPF (Open Shortest Path First) - widely used in the internet.
27
Difference of link-state routing from distance vector?
In link state routing each router floods a link state packet listening its neightbours and the cost of the each link to and all other routers. Each router can make theyre own network topology path.
28
How is the flooding of LSPs managed to avoid loops
- Use sequence numbers
- Discarding older LSPs
- Include TTL
- Soft state approach
29
Advantages of link state routing
- Link has better convergence properties
- Link state is robust with faulty of malicious routers
30
Broadcast routing
Want to send packets to all other hosts in network.
31
Reverse Path Forwarding (RPF)
A router only forwards a packet if it arrives on the same link that the router would use to send a packet back to the source.
32
Spanning Tree
- Forward copies over a tree rooted at the source host that includes other hosts
- Each router now only gets one copy of each broadcast packet
- no dupes
33
How do you determine a spanning tree?
Each router forwards packets only to neightbours that route to the source via it.
34
Anycast routing
Any one of several possible destinations is acceptable.
35
How is anycast implemented
Multiple hosts share the same IP Address, routers forward the nearest one
36
Multicast Routing
Sending packets to a selected group of hosts, termed multicast group
37
What does multicast routing require
- Group management protocol
- Routing protocols
38
Why is multicast more complex than broadcast
It must only reach group members and groups can dynamically change
39
Problem of consturcting a minimum-weight Steiner Tree with Multicast?
Can change dramatically when the set of routers are changed. Even slightly can mess wit hthe roral weight of the links in the trees.
40
Core-Based Tree
Group-shred trees rooted at a core used for distributing multicast traffic. Group-shared tree. One main controller at the center
41
How is a core-based tree constructed?
Requires routers with local hosts belonging to the multigroup to send "join: messages towards a core.
42
Source Specific Tree
Tree rooted at the source and splanning only to those routers with group memebers
43
MOSPF
Is a mutlicast extension for Open Shortest path protocol. Computes soure-specific trees using full topology graph; not used in practice
44
Pruning in reverse path forwarding
Routers that dont need group send a prune message upstream removing branches of the tree. Used in DVMRP. Not used anymore
45
Source-Based Core-based tree
Each source becomes a core, recieves must send joins to each source.
46
PIM-SM Hybrid Approach
Start with core-based tree, group shared, then build specific soure-specfic trees for active sourves. USED IN INTERNET Right now
47
When the congestion occur
Too much traffic in some portion of network
48
Consequences of Congestion Control
- Dropped Packets
- Long Delays
- Unfairness
- inefficent use of resources
49
Approaches to fixing Congestion Control
- Network provisioning
- Routing Change
- Admin Control
- Traffic Throttling
- Load Shedding
50
Two Types of traffic throttling
End-to-end
Network Assisted
51
End-to-end traffic throttling
No feedback from network; congestion is inferred from obsered loss or dely by communicating enpoints.
52
Network Assisted Approaches to traffic throttling
Routers explicit feedback to traffic soures either directly or indirectly via receviers.
53
Two Network Assisted Approaches
- Directly: "Choke" packets sent by congested routers back to high rate sources
- Indirectly: Congested routers set a bit or fill in a field, in the packet header of selected packets transiting through. Reciving host than returns this feedback to source host.
54
Bufferbloat problem
Overly large buffers cause long delays before packets are dropped, hurting real-time traffic
55
ECN (Explicit Congestion Notification)
Routers mark packets, recievers signal congestion to senders without packet loss, Used in data centers
56
Load Shedding
Routers drop packets to reduce congestion using either tail drop or active queue management
57
AQM Active Queue Management
Routers begin dropping or marking packets early, before buffers are full
58
PI Controller in AQM
PI = proportional–integral
A method that updates the drop probability based on both delay offset and rate of change of delay
59
What is the internets default network layer model regarding Quality of Service
Best Effort - no guarantees on dely, throughput, or packet loss
60
Quality of Service importance for multimedia
- continuous media
- Real time support
- better scale
- Approximability
61
Two approaches to supporting multimedia applications
- Application-layer adaptation
- Network Layer QoS support
62
General Quality of Service concerns
- Fair resource sharing
- ISP service level agreements with users
- Net Neutrality Concerns
- Premium Service Plans
63
Three Levels where Quality of Service might be applied
- End-to-end flow (video stream)
- Aggregate flows (business to ISP)
- Traffic class (VoIP)
64
Admission Control in Quality of Service
Process where flow requests network access with its QoS needs, and the network rejects or accepts based on avaliable resources
65
Traffic Shapping
Delay packets to contorm to the rate; regulate average rate and burstiness of flows.
66
Traffic policing
Drops packets that exceed the allowed rate limit.
67
Token Bucket Algorithm
Traffic Shapping and policing, Wanting to enter entwork needed one "token" for each byte
68
In token bucket, what do R and B represent
R - Token Generation Rate (bytes/sec)
B - Bucket Size
69
In token bucket how is duration S calculated.
S = B / M - R
M = LINK RATE, R = TOKEN RATE, B = BUCKET SIZE
70
Packet Scheduling
Allocate router/link resources to packets based on flow, flow aggregate or traffic classes based on QoS policies
71
What scheduling has the problem of starvation
Strict Priority Queueing
72
Round-Robin Scheduling
Each flow gets equal turns in a cycle
73
Weighted fair queueing
Each flow gets service proportional to its weight
74
Integrated Services
Per-flow resources reservations for individual flows using RSVP Signaling.
75
Key drawbacks of Integrated services
- Requires support on all routers
- Requires per-flow state
- Poor Scalability
76
RSVP in Intregated Services
Resource Reseveration Protocol; reciever sends signaling message to sending asking routers to reserve resources
77
Differentiated Services
A class-based Quality of Service approach where routers forward packets bassed on traffic class markings set at the network edge.
78
Per-Hop Behaviors (PHBs)
Router-defined policies for how each class of traffic is treated
79
How many traffic classes are possible in differentiated services
64 (6-bit field in IP Headers)
80
Two main standardized service types in differentiated services
- Expedited Forwarding
- Assured Forwarding
81
Major Limitation of differentiated services
Packet markings are not guaranteed to be honored across different ISP Networks
82
IPv4
- 20 Byte Fixed-length portion of header
- Variable length portion of header for optons
- Payload
- version (4)
- Header length
- DiffServ (6)
- ECN (2)
- Length (16)
- Identifier (13)
- TTL (8)
- Upper Layer protocl (8)
- Header checksum (16)
-
83
Purpose of TTL field in IPv4
To prevent routing loops, decremented each hop; if 0, drop packet and send an ICMP bacl
84
Header checkum in IPv4
Detects errors in the header using 1's complement arithmetic
85
MTU Maximum Transmission Unit
Max IP Datagram size supported by a network/link
86
Process of fragmentation
Wehn a router recieves a datagram which size exceeds the MTU of out the outgoing link it must be forwarded. The router can fragment the datagram into smaller segments.
87
Who reassembles fragments in IPv4
The Destination Host
88
What happens if one fragment is lost
The Entire Datagram is discarded; Then resend
89
Path MTU Discovery
The source sets DF (Dont fragment) flag to one. if the datagram too big, the router will drop it and send a ICMP message back. Assuming its recieved the source can adjust the max size of datagrams it is sending
90
How many bits are in IPv4 address
32 bits, written in dotted decimal
Each Network Inteface has its own IP Address
91
Classful IP Addressing
32-bit address space divided in 5 classes with specific sage and subnet mask. It lacked flexibility and could not scale
92
What has replaced classful addressing
CIDR (Classless interdomain routing)
93
CIDR - Classless Interdomain Routing
200.23.16.0/20 || Used nowadays
94
What rule is used when multiple CIDR prefixed match?
Use the longest prefix match
95
What do we use to make lookup for CIDR fast
TCAM: Ternary Content-Addressable Memory
96
When does a router use destination hosts layer 2 address
When a router is directly connected to the destination layer 2 network
97
When does a router use next-hop routes Layer 2 address
When the router is not directed connected ot the destination network and must forward the packet to the next router
98
What protocol is used to map IP addresses to Layer 2 Addresses
ARP (Address, Resolution Protocol)
99
What if ARP table entry is missing
Router broadcasts an ARP query on the network
100
Who allocates large IP address blocks globally
ICANN, delegates blocks to Reginal Internet Registries
101
Why do routing tables grow so large
Due to fragmented IP Address allocations with reduce route aggregation
102
Longest pre fix match rule
A routers forwards a packet based on the route with the most specific matching prefix
103
Two main types of assigning IP Addresses to hosts
Manual Configuration (Static)
DHCP (Dynamic)
104
What protocol lets hosts dynamically obtain IP Addresses
DHCP (Dynamic host configuration protocol)
105
How deos a host communicate with a DHCP Server
Using a layer 2 broadcast with 0.0.0.0 IP field
106
4 Basic DHCP Messages
Discover
Offer
Request
ACK
107
What more can a DHCP Offer than a IP Address
Default Gateway IP
DNS Server IPs
Subnet Mask
Lease Duration
108
NAT Network Address Translation
The process that maps IP addresses from private network to a public IP address
109
Motivations for Network Address Translation
- Desire to conserve IPv4 address with NAT
- Tiered Pricing
- Security with NAT hosts cannot be visible to the outside work
110
How does a NAT router handle an outgoing connection
- Creates a translation table entry mapping local IP port to external IP port
- Rewrites source IP/Port on outing datagrams
- Rewrites dest IP/port on replies
111
Limitation of basic NAT for inbound traffic
Host inside NAT cannot be reached directly unless special configurations are made; port forwarding
112
What if a host in a private networks wants to be accessible to the outside world
- Manual config of the NAT router
- Use of protocl allowing hosts to dsicover and config NAT router
- "Hole punching" made possible by the fact many NAT will use same external port mapping for all packet flows
113
What is hole punching in NAT Traversal
A technique where two hosts beind NAT use known external port mappings to send simultaneous messages to each other, tricking NAT routers to open inbound paths
114
What is an Autonomous System (AS)
A network or group of networks under a single administrative control, identifed by an autonomous system number ASN
- Internet is a confederation of networks
115
Three main types of AS Relationships
- Stub Networks: carry only traffic that is sourced to/from an endpoint
- Peer-to-peer
- Customer-provider
116
IXP (Internet Exchange Point)
A physical facility where Autonomous Systems interconnect to exchange traffic directly.
117
Difference between Intra-AS routing and Inter-AS routing
Intra-AS Routing: Within a single AS and is performance-based
Inter-AS routing is between AS's and is policy based
118
Standard inter-AS routing Protocol?
BGP (Border Gateway Protocol)
119
BGP Border Gateway Protocol
is primary routing protocl used on the internet to exchange routing information between different autonomous systems.
120
hot potato routing
Choosing the closest exit point to enter another AS to minimize internal traffic load
121
Border Gateway Protocol Hijacking
When an Autonomus System Illegitimately annouces a route to a prefix it doesnt own diverting traffic.
122
Route Origin Authorization
Specifies which AS is allowed to advertise an IP Prefix usinf RPKI
123
Within an AS what is PIM-SM
Protocol Indpendent Multicast is used to contruct multicast distribution trees
124
How are multicast groups formed in IP Networks
Using class D addresses, hosts use IGMP to inform local membership in multicast groups
125
What does IGMP do?
Internet Group Management Protocol; Allows a host to notify its lcoal router that it wants to join or stay in the multicast group Periodically broadcasts a query that must be responded too
126
MPLS ("Multiprotocol label Switching")
It introduces virtual circuit routing at the level of flow aggreates to improve traffic management and QoS
127
How does MPLS forward packets?
By utilizing a short label in the MPLS header instead of the full IP address these labels are swapped at each hop based on local tables
128
Label Stacking in MPLS
Supports multiple nested labels allowing for multiple levels of abstraction
129
What is Software Defined Networking
SDN seperates the control plane (decision making) from the data plane (packet forwarding) to centralize control logic on controller machines
130
Flow tables in SDN
Tables in switches that contain match-action rules for forwarding or processing pakcets on header field value
131
OpenFlow
widely adopted protocol that allows SDN controllers to communicate with and control switches
132
IPv6
Created to solve the IPv4 address exhaustion problem by expanding the address space to 128 bits
133
Key difference between IPv6 and IPv4
IPv6 elimates the header checksum and router-side fragmentation to reduce router processing overhead
134
IPv6 Datagram fields
version - 4 bits
diffserv - 6 bits
ECN - 2 bits
payload length - 16 bits
next header - 8 bits
hop limit - 8 bits
source address, dest address - 128 bits
135
IPv6 Flow Label
To mark packets that belong to the same flow so they can recieve special Quality of service treatment
136
what is NAT64
A network address translation to allow IPv6 to talk to IPv4 servers
