Download presentation
Presentation is loading. Please wait.
Published byWilfrid Glenn Modified over 7 years ago
1
Network Layer Datagram vs. Virtual circuit Routing Congestion control
Internetworking Internet Example Systems Ying-Dar
2
Datagram vs. Virtual Circuit --Two alternative for network layer
Datagram (connectionless): Each packet passes through the network as independent entity with full destination address; routing decision is based on the routing table. Virtual circuit (connection-oriented): Full duplex virtual channel is established by creating an entry in virtual circuit table on each node along the path; based on another routing table. Virtual circuit ID defined packet destination address (abbreviated addressing); routing decision is based on the virtual circuit table. Ying-Dar
3
Datagram Approach B C A D F E D C 2 F D B 3 packet: dest src ... data
Dest Next Delay (Routing Table) packet: dest src ... data Ying-Dar
4
Virtual Circuit Approach
C H C 1 H 1 A A 1 F B D H 2 F 1 B 1 D 1 (Routing Table) F H E H E 1 D 2 D B 3 B C C H C 1 H 1 A D F H 2 F 1 H 3 host H B C 2 F E F H 1 E router B E 1 H 2 B 1 A F E D H 3 E 2 A 1 H B D 1 A 2 C B 1 H incoming outgoing A 3 C 1 D B (Virtual Circuit Table) V.C. ID packet: 2 data 1 data data 1 data (at A) (at B) (at F) (at D) Ying-Dar
5
Comparison of Datagram and Virtual Circuit
Robustness to failures more less Overhead call setup memory for VC table reassemply Example Internet ATM Ying-Dar
6
IP vs. ATM datagram subnet vs. virtual-circuit subnet
soft-state flow vs. hard-state connection Applications Application Q.2931 IP-over-ATM LAN Emulation TCP UDP RSVP AAL IP ATM Ethernet Ether Switch HDLC ... Sonet connection subnet connection-oriented subnet connection-oriented socket connection-oriented APIs connectionless socket Ying-Dar
7
Routing Routing policies Shortest path routing Flow-based routing
Flooding Distance vector routing Link state routing Hierarchical routing Mobile routing Broadcast routing Multicast routing Ying-Dar
8
Routing Policies Static vs. dynamic routing
Centralized vs. distributed routing Building blocks --network status monitoring (local, global, partial) --network status reporting (exchange, broadcast) --route computation (delay, bandwidth) --route implementation (routing table, virtual circuit table, route cache) Ying-Dar
9
Shortest Path Routing Static and centralized Dijkstra Algorithm (1959)
1. To compute the distance (cost) from A to all nodes, label all other nodes as (,–) and tentative, while marking A as permanent and working node X. 2. Re-label the nodes adjacent to X as (distance to A,X) if X provides a shorter path to A. 3. From the tentative nodes, mark the one with the smallest label permanent and new working X. Go to step 2 until all nodes are permanent. Ying-Dar
10
Shortest Path Routing (Cont.)
B (2,A) C (9,8) 7 distance to A next hop 2 2 3 3 (4,8) A E F (6,E) D (10,H) 2 6 1 2 2 G (5,E) 4 H (8,F) Ying-Dar
11
Flow-based Routing Static and centralized
Consider topology (connectivity and capacity) and traffic matrix to allocate traffic flows over the topology to optimize performance (delay, loss, blocking) Optimization techniques exist in routing the flows. Ying-Dar
12
Flooding Flood the incoming packet on every outgoing link except the one it arrived on. Two ways to limit the scope of flooding: --hop count in packet header --sequence number in packet header to avoid duplicate flooding Ying-Dar
13
Distance Vector Routing
Also named as distributed Bellman-Ford, original ARPANET, RIP Distance vector exchange: Periodically each node sends the distance vector to its neighbors. Routing table update: Di(k) = min [ Di(j) + Dj(k)], j is i’s neighbor Ri(k) = j*, which is i’s neighbor that minimize Di(k) Initialization: Di(k) = 0 if k=i if kI Ri(k) = Ying-Dar
14
Distance Vector Routing (Cont.) (For destination F)
(,) (,) (11,C) (9,D) (9,D) 7 2 (,) (4,F) (4,F) (4,F) (4,F) B C 1 3 1 4 3 2 2 2 F (0,F) A D (,) (12,E) (11,B) (DA(F),RA(F)) 6 1 5 3 3 E 7 (,) (7,F) (7,F) (7,F) (7,F) Ying-Dar
15
Problems with Distance Vector Routing
Not taking link bandwidth into account (only considering instantaneous queue length) instability and oscillation Taking a long time to converge good news: travel at the rate of one hop per exchange bad news: count-to-infinity (because no router ever has a value a few more higher than the minimum of all its neighbors) Ying-Dar
16
Link State Routing Examples: IS-IS, OSPF,
Learn about neighbors and their network address ( HELLO packet ) Measure link state ( ECHO packet ) Build link state packets ( router id, sequence, age, ( neighbor, cost ),.... ) Distribute link state packets to all other routers --check and update the table ( source router, sequence, age, send flags, ACK flags ) Compute new routes ( run Dijkstra’s algorithm locally ) Ying-Dar
17
Hierarchical Routing A router has one entry, in its routing table, for each router in the same region, and also one representation entry for each of other regions. Example: For a subnet with 720 routers partitioned into 24 regions of 30 routers each, each router needs 53 entries ( 30 local + 23 remote ). For a subnet with n routers, optimal number of hierarchical levels is ln(n) and number of entries per router is eln(n) Ying-Dar
18
Mobile Routing Mobile ( wired or wireless ) hosts travel without changing their network address. Need home agent at home site and foreign agent at remote site. Mobile hosts registers to foreign agent when travelling to a foreign LAN or cell. Foreign agent informs home agent of the mobile hosts. Packets sent to the mobile host are intercepted by home agent and tunneled to foreign agent. Home agent informs the packet sender foreign agent’s address. Subsequent packets to mobile host are re-directed to foreign agent. Ying-Dar
19
Broadcast Routing Possible methods: flooding, multi-destination routing, optimal sink tree, reverse path forwarding Reverse path forwarding: approximate the optimal sink tree ( router checks to see if the packet arrived on the line that is normally used to send packets to the source of the broadcast ) I B • C • B • C • F H J N A • D • A • D • F • E • E • F • A D E K G O M O I • G • I • G • E C G D N K H • H • • L • N • J • L • N • J H B L • K • K • M • O • M • O L B Ying-Dar
20
Multicast Routing To do multicast routing, two approach exist: --pruned spanning tree per source per group --core-base tree per group Pruned spanning tree per source router per group 1.with link state routing, pruned locally at source router 2.with distance vector routing, PRUNE message stop multicasting Core-base tree per group --send multicast packets to the core ( root ) of the group Ying-Dar
21
Congestion Control Congestion phenomena
Policies that affect congestion Classification of congestion control: best-effort and guaranteed Choke packet scheme Sliding window-based flow control Rate-based flow control Weighted fair queueing Admission control Traffic shaping and policing Ying-Dar
22
Congestion Phenomena Throughput Offered Load 22 Ying-Dar Lin@CIS.NCTU
capacity of subnet ideal Throughput controlled uncontrolled deadlock Offered Load Ying-Dar
23
Policies That Affect Congestion
Ying-Dar
24
Classification of Congestion Control
Ying-Dar
25
Choke Packet Scheme When queue length or utilization exceeds a threshold, the router sends a choke packet back to (1) the source host or (2) the immediate upstream router, giving it the destination found in the affected packet stream. Case (1): source choke packet slower response Case (2): hop-by-hop choke packet higher buffer requirement Ying-Dar
26
Sliding Window-based Flow Control
To limit the number of outstanding ( un-acknowledged ) packets, a fixed or variable window is allocated for a packet stream ( flow ) between a pair of source and destination. Only when the left-most packet is acknowledged can the window be slided forward. Window size depends on bandwidth, propagation delay, traffic load. Ying-Dar
27
Rate-based Flow Control
When queue length or utilization reaches a certain value, router/switch (1) computes the allowed rate for the source of each packet stream, or (2) informs the source the congestion status. Case (1): explicit rate flow control Source sets its rate to the assigned allowed rate. Less oscillation Complex router/switch computation Case (2): congestion notification flow control Source adjusts its rate according to some rule, e.g. linear increase, exponential decrease larger oscillation simpler router/switch Ying-Dar
28
Admission Control For a flow or connection that requires guaranteed service, admission control is exercised to determine the grant/reject, depending on the traffic descriptor performance parameters of the flow or connection. Traffic descriptor ( Tspec ): peak rate, mean rate, burst length Performance parameters ( Rspec ): delay, loss, jitter Request for the flow or connection is granted if network resource, bandwidth and buffer, is available along a path to accomodate the flow or connection. Ying-Dar
29
Traffic Shaping and Policing
Leaky Bucket scheme can be used to (1) smooth/enforce a flow or connection, at host side, to conform to the declared traffic descriptor, or (2) police a flow or connection, at network-side, to regulate the input traffic. token : token arrival rate (mean rate) : token bucket size (burst length) packet/cell packet/cell c: link capacity (allocated bandwidth) b :buffer size (allocated buffer) Leaky (Token) Bucket Queue Ying-Dar
30
Packet Scheduling At routers/switches, certain queueing discipline is enforced to schedule (or arbitrate), for each output line, which input stream to serve next. Method (1): mixed single queue for each output line --unfairness, does not take connection or flow bandwidth into account Method (2): fair queueing (multiple queue for each output line, one for each source) packet-by-packet round robin byte-by-byte round robin weighted byte-by-byte round robin (weight can the number or total bandwidth of the flows/connections from the source) Ying-Dar
31
Internetworking Scenario, Devices, Approaches
Scenario: LAN--LAN LAN--WAN--LAN WAN--WAN Devices: Repeater, Bridge, Router, Gateway ( Higher-level internetworking devices are more capable to handle and convert the difference in interconnected networks ) Approaches: Connection-oriented: concatenated virtual circuit Connectionless: through different networks, different routes Tunneling: special approach for LAN-WAN-LAN where two LANs are homogeneous and WAN can be connection-oriented or connectionless Ying-Dar
32
Internetworking Local and Wide Area LAN Interconnection
Local area LAN interconnection: bridge, router, switch Wide area LAN interconnection: modem, leased lines, ISDN X-25, Frame Relay, SMDS B-ISDN, Internet Ying-Dar
33
Internet Internet protocol hierarchy IP header
IP address and subnetting Control protocols: ICMP, ARP, RARP, BOOTP Routing protocols: OSPF and BGP IGMP, Mobile IP, CIDR, IPv6, RSVP Ying-Dar
34
Internet Protocol Hierarchy
OSPF BGP SMTP NNTP HTTP Telnet FTP NFS YP Mount DNS BOOTP RPC TCP UDP ICMP ARP RARP IP Ying-Dar
35
Options (0 or more words)
IP Header 32 Bits Version IHL Type of service Total length Identification D F M F Fragment offset Time to live Protocol Header checksum Source address Destination address Options (0 or more words) Option Description Security Specifies how secret the data gram is Strict source routing Gives the complete path to be followed Loose source routing Gives a list of routers not to be missed Record route Makes each router append its IP address Timestamp Makes each router append its address and timestamp Ying-Dar
36
IP Address and Subnetting
32 Bits class Range of host address A Network Host to B 10 Network Host C 110 Network Host to D 1110 Multicast address to E 11110 Reserved for future use to to This host … Host A host on this network Broadcast on the local network Network … Broadcast on a distant network 127 (Anything) Loop back class B add entries of (this-network, subnet, o) and (this-network, this-subnet,host) to routing table 10 network subnet host subnet mask (e.g.) Ying-Dar
37
Internet Control Protocols
ICMP ( Internet Control Message Protocol ) Destination Unreachable, Time Exceeded, Parameter Problem, Source Quench, Redirect, Echo Request, Echo Reply, Timestamp Request, Timestamp Reply, etc. ARP ( Address Resolution Protocol) Symbolic name Domain Name System IP address ARP MAC address RARP (Reverse Address Resolution Protocol) server --through local broadcast IP BOOTP server ---- through UDP Ying-Dar
38
OSPF (Open Shortest Path First) Interior Gateway Routing Protocol for AS (Autonomous System)
Requirements: open, variety of distance metrics, dynamic, based on type of service, load balancing, hierarchical security, tunnel allowed. Each AS, divided into areas, has a backbone area connecting all other areas, by point-to-point links, broadcast links, or tunnels in WAN. Three kinds of routes: intra-area: link state shorting path routing inter-area: (1)from source to backbone, (2)across backbone to dest area, (3)to dest inter-AS: BGP (Border Gateway Protocol) - exterior gateway routing protocol OSPF messages: Hello, Link state update, Link state ack, Database description, Link state request Ying-Dar
39
BGP (Border Gateway Protocol) Exterior Gateway Routing Protocol for Inter-AS
Designed to allow many kinds of routing policies to be enforced in the inter AS traffic, by route scoring function. A distance vector protocol where BGP router keeps track of the exact path used in routing table and tells its neighbor the exact path, instead of cost, it is using. Pairs of BGP routers communicate with each other by establishing TCP connections. No count-to-infinity problem Ying-Dar
40
IGMP (Internet Group Management Protocol)
Permanent and temporary group addresses Permanent groups: (e.g.) all systems on a LAN all routers on a LAN all OSPF routers on a LAN all designated OSPF routers on a LAN Temporary groups: IGMP query: each multicast router multicasts to hosts on its LAN to ask them the groups their processes belong to IGMP response: each host responds class D address it is interested in Each multicast router constructs a pruned spanning tree per group, using a modified distance vector protocol. Tunneling is used in the pruned spanning tree to bypass nodes not in a spanning tree. Ying-Dar
41
CIDR (Classless InterDomain Routing)
More than half of all class B networks have fewer than 50 hosts. Too many small class C networks would enlarge routing table dramatically. CIDR: allocate the remaining class C networks (about 2 million) in variable-sized blocks ~ : for Europe (194... and 195... entries) ~ : for North America ~ : for Central and South America ~ : for Asia and the Pacific ~ : reversed for future use Example: NCTU asks for 4096 addresses and is assigned through along with mask If a packet is addressed to : mask base address next . . . ... AND . . . match Ying-Dar
42
IPv6 To address three major problems:
address space, encryption and authentication, flow label for QoS treatment 32 Bits Version Priority Flow label Payload length Next header Hop limit Source address (16 bytes) Destination address (16 bytes) Ying-Dar
Similar presentations
© 2025 SlidePlayer.com Inc.
All rights reserved.