CS 455 Internet Architecture, Page 3 ffl By 1985, the ARPANET was heavily used and congested; the National Science Foundation (NSF) initiated the NSFN

Transcription

1 CS 455 Internet Architecture, Page 1 Network Layer, Part 3 Internet Architecture These slides are created by Dr. Yih Huang of George Mason University. Students registered in Dr. Huang's courses at GMU can make asingle machine-readable copy and print a single copy of each slide for their own reference, so long as each slide contains the copyright statement, and GMU facilities are not used to produce paper copies. Permission for any other use, either in machine-readable or printed form, must be obtained from the author in writing. CS 455 Internet Architecture, Page 2 History ffl U.S. DoD Advanced Research Project Agency (DARPA) created ARPANET starting ffl ARPANET was the first wide-area general purpose packet network, severing universities and ARPA contractors. ffl In early 70's, the concept of internetworking was advanced. Most networks are established by independent entities. Bridged networks require close interaction between owners and are limited to local areas. Users soon demanded universal connection for computer networks, as is provided by the phone network. The ARPANET became the core of the Internet experiment.

2 CS 455 Internet Architecture, Page 3 ffl By 1985, the ARPANET was heavily used and congested; the National Science Foundation (NSF) initiated the NSFNET as the new backbone of the Internet. original NSFNET links were 56K, updated to T1 (1.544 Mbps) in 1988 and later to T3 (45 Mbps) in 1991 the NSFNET was decommissioned in April ffl The contemporary Internet does not have an official backbone network; instead Network Access Points (NAPs) were established for multiple backbone networks to exchange traffic: 1. Spring NAP Pennsauken, NJ 2. PacBell NAP San Francisco, CA 3. Ameritech Advanced Data Services NAP Chicago, IL 4. MFS Datanet Washington, D.C. ffl Each NAP is essentially a high speed LAN, using FDDI or ATM technologies. CS 455 Internet Architecture, Page 4 TCP/IP The Internet employs the TCP/IP protocol suite, which includes 1. Internet Protocol (IP) ffl unreliable, connectionless datagram delivery ffl routing 2. User Datagram Protocol (UDP) ffl unreliable, connectionless ffl allow processes to send datagrams to one another 3. Transmission Control Protocol (TCP) ffl full-duplex, reliable transport service ffl flow and congestion control ffl stream delivery service

3 CS 455 Internet Architecture, Page 5 IP Addresses ffl 32-bit addresses ffl Class A: 128 nets with 16M hosts each 0nnnnnnn hhhhhhhh hhhhhhhh hhhhhhhh ffl Class B: 16K nets with 64K hosts each 10nnnnnn nnnnnnnn hhhhhhhh hhhhhhhh ffl Class C: 4M nets with 256 hosts each 110nnnnn nnnnnnnn nnnnnnnn hhhhhhhh ffl Class D: multicast addresses, 256M groups ffl Class E: reserved for future use 1110gggg gggggggg gggggggg gggggggg 11110xxx xxxxxxxx xxxxxxxx xxxxxxxx CS 455 Internet Architecture, Page 6 IP Number Notation A single decimal notation is used by humans, regardless the class of the address. Binary: Examples: Decimal: ffl cs.gmu.edu = ffl site.gmu.edu = ffl bacon.gmu.edu =

4 CS 455 Internet Architecture, Page 7 What Does an IP Address Mean? ffl There is no 1-to-1 correspondence between IP addresses and nodes (hosts or routers). ffl Instead, IP addresses identify interfaces. CS 455 Internet Architecture, Page 8 Domain Name System (DNS) ffl Of course, we humans don't even want to memorize decimal numbers; we use names. ffl The DNS is like a directory hierarchy: you start with a top-level domain and specify sub-domain name and sub-sub-domain name, and so on, in a right-to-left manner. ffl There are three types of top-level domains. 1. arpa is a special domain name used for address-to-name mappings (not discussed further) 2. seven 3-character domains, called generic domains com, edu, gov, ini, mil, net, org 3. 2-character country codes, according to ISO 3166

5 CS 455 Internet Architecture, Page 9 ffl Top-level domain names are maintained by the Internet Cooperation for Assigned Names and Numbers (ICANN). ffl Second level domain names (i.e., can be obtained from one of the accredited registrars listed by ICANN. ffl Management of lower-level names is delegated. For exame, GMU is responsible for managing the domain gmu.edu. GMU may further delegate the responsibility of cs.gmu.edu to the CS department. CS 455 Internet Architecture, Page 10 DNS Servers ffl A hierarchy of DNS severs are used to translate domain names to IP addresses. ffl Each host is manually configured with the IP address of a local DNS server. ffl When the local DNS server cannot resolve a request, it contacts a higher-level sever. ffl Each authority of a domain maintains a primary name server and optionally several secondary name servers. the primary name server is maintained by humans, while a secondary server obtains its database from the primary server secondary servers are used to reduce the workload of the primary server

6 CS 455 Internet Architecture, Page 11 ffl A group of root name servers maintain all top-level domain names; each root server contains a complete mapping between top-level domains and their primary server addresses ffl For example, when mulder.gmu.edu requests the IP address of scully.stanford.edu, the following DNS severs will be involved: 1. a (primary or secondary) server of GMU 2. a root server 3. the primary server of Stanford CS 455 Internet Architecture, Page 12 IP Header Vers HLen Service Type Total Length Identification Flags Fragment Offset Time to Live Protocol Header Checksum Source IP Address Destination IP Address IP Options Padding Data ffl Vers: protocol version (current=4, new=6) ffl HLen: Header length in 32-bit words (max=60 bytes)

7 CS 455 Internet Architecture, Page 13 ffl Service Type: precedence, delay-sensitive, reliability-sensitive,... mostly not used but may beimportant in the future ffl Total Length: packet length in bytes (max bytges) ffl Identification, Flags, Fragment Offset: used in fragmentation, i.e., passing a packet through a network where its length except the network's Maximum Transmission Unit (MTU). ffl Time to Live (TTL): maximum number hops the packet can survive; set by the sender and decremented by each intermediate router. One application of TTL is to restrict the damage of routing loops. ffl Protocol: code for transport protocol ffl Header Checksum: error check for the header uses exclusive-or, rather than CRC easy for incremental update of CS 455 Internet Architecture, Page fragment offset 2. time to live ffl Source/Destination IP Addresses ffl Options 1. Security, not in use now 2. Record route: each intermediate router records its IP address in the options data area 3. Time stamp: each router records the time 4. Loose source route: the sender lists a series of routers that must be visited by datagram in the specified order; other routers may bevisisted when the packet moves from one listed router to the next 5. Strict source route: only listed routers can be visited.

8 CS 455 Internet Architecture, Page 15 Fragmentation ffl Datagrams transmitted across the physical network in frames. ffl Each network usually imposes a maximum transmission unit (MTU); the MTUs along the delivery path of an IP datagram may vary. ffl When an IP datagram of b-byte long arrives at a router and is about to go across a network with MTU less than b, the router divides the datagram into fragments a prior to transmission. each datagram from a sender to a recipient has a unique Identification. this Identification is copied to every fragments of the datagram the Offset field contains the offset of the first byte of the fragment within the entire datagram CS 455 Internet Architecture, Page 16 a more-fragment bit in the Flags field is turned on for all fragments except the final one besides these, fragments are treated like regular datagrams: each of them contains complete source and destination addresses and is routed independently ffl Re-assembly of the datagram takes place only at the destination. this is true even if all subsequent networks can transmit the entire datagram in one frame

9 CS 455 Internet Architecture, Page 17 Address Resolution Protocol (ARP) ffl Consider what happens when the packet arrives at the network, say an Ethernet, in which the destination host resides. precisely, the packet arrives at a router that is attached to the network recall that the packet carries the destination IP" address however, the router needs the destination Ethernet" address so it can deliver the packet to the destination. ffl In essence, the router broadcasts an ARP request message via the Ethernet, asking the owner of the designated IP address to respond with its Ethernet address. CS 455 Internet Architecture, Page 18 ffl Improvements? ffl ARP is also used when two hosts connects to the same network communicates with the TCP/IP protocol suite.

10 CS 455 Internet Architecture, Page 19 IP Routing ffl Direct Delivery sender determines that destination is on a local network by checking the netid in its IP address against the netid of the destination if the network is a LAN, the ARP is used to obtain to obtain the physical address of the destination a data-link frame is composed and transmitted ffl Indirect Routing sender forwards the packet to a router that is attached to the network the router is responsible for forwarding the packet toward its detination CS 455 Internet Architecture, Page 20 Routing Tables The routing table of a router comprises a list of network-id to interface mappings. ( , 0) ( , 1) (160.32, 2) ( , 2) Host Router 2 ATM Switch

11 CS 455 Internet Architecture, Page 21 Internet Routing Hierarchy ffl Due to its scale, no single routing protocol can take onthe entire Internet. ffl The Internet is divided into Autonomous Systems (AS). ffl Each AS comprises the routers and networks governed by a single administration authority. Examples: GMU, an IP service provider, Ford, FBI, etc. ffl Routing happens at two levels: Intra-AS routing, also called internal gateway routing Inter-AS routing, also called external gateway routing CS 455 Internet Architecture, Page 22 Two-Level Routing of The Internet the AS of the backbone ISP GMU AS site unix Other AS Your home PC the AS of your ISP Gateway Router Regular Router

12 CS 455 Internet Architecture, Page 23 ffl If the destination of a given packet is inside the source AS, then the packet is routed by the intra-as routing protocol of the AS. RIP and OSPF are the two most popular intra-as routing protocols. ffl If the destination is outside the source AS, then 1. the packet is first routed to a boarder gateway router of the AS, using intra-as routing. a broader gateway router is a router in the AS that connects to at least one router in another AS such a router performs both intra-as and inter-as routing CS 455 Internet Architecture, Page The broader routers of all ASes use an inter-as routing protocol to decide a series of ASes to deliver the packet. EGP (External Gateway Protocol) is the early choice for this task; this protocol has been abandoned. BGP is the current inter-as routing protocol. 3. An intra-as routing protocol is used to route the packet across each intermediate AS and, when the packet reaches the destination AS, to deliver the packet to the destination host.

13 CS 455 Internet Architecture, Page 25 The Address Depletion Crisis ffl We are running out of IP addresses. ffl However, it is not that we have used up all 2 32 addresses. it is class B addresses that are in shortage since class A networks are too large and class C networks are too small, large portions of the address space are not very useful ffl Solutions: 1. dynamic address assignment 2. classless interdomain routing (CIDR) 3. private addresses and address translation 4. next generation IP CS 455 Internet Architecture, Page 26 Dynamic Address Assignment ffl Each ISP is allocated a block of IP addresses. ffl An IP address is assigned to your PC only when you dial up to your ISP; the address is released after you hang up. ffl This reduces the number of IP addresses to the number of on-line users, rather than all the users of the ISP. ffl Most ISPs have stopped assinging static IP addresses to customers.

14 CS 455 Internet Architecture, Page 27 Classless Interdomain Routing (CIDR) Using a mask to explicitly indicate the length of network ID field Host , 28 Router ATM Switch , , , 27 CS 455 Internet Architecture, Page 28 Private Addresses and Address Translation ffl Some IP addresses, called private addresses, are reserved for internal use in private networks; public routers will not forward packets belonging to these addresses to to to ffl A corporation can use private addresses at its own discretion. the same addresses may also be used by other corporations ffl To access the Internet, the corporation obtains a small set of globally unique IP addresses.

15 CS 455 Internet Architecture, Page 29 ffl When a machine inside the corporation communicates with outside world, its private IP address is translated to a globally unique IP address by a border router. ffl This binding between private addresses and public addresses is dynamic. ffl This method works well if at any given moment onlya small portion of the machines communicate with the outside world. CS 455 Internet Architecture, Page 30 The Next Generation IP ffl usually called IPng or IPv6 (current generation is IPv4) ffl uses 128-bit addresses ffl simplified header format for fast forwarding ffl practical issues: When will people start using IPv6? How do we manage the transition?