Internet Application Theory Applications Internet Application Ibrahim Otieno

Internet Application Theory & Applications

Internet Application - Ibrahim Otieno iotieno@uonbi. ac. ke +254 -0722 -429297 SCI/ICT Building 2 nd Floor Rm. 201

Binding Protocol Addresses

Protocol Addresses &Packet delivery Application program generate packets to send across internet, software place data in packet that contains destination address Software in each host / router uses destination address to select a next hop for packet Once next hop selected, software transfers packet across network to selected host/ router To provide illusion of one large network, s/w works with IP addresses to forward packets

Protocol Addresses &Packet delivery Protocol addresses can’t be used to transmit across physical hardware Hardware doesn’t understand IP addressing Frame sent across physical network must use hardware’s frame format Addresses in frame must be h/w addresses

Protocol Addresses &Packet delivery Protocol address for next hop must be translated to equivalent hardware address before a frame can be sent In summary, protocol addresses are abstractions provided by software; Network hardware doesn't know to locate a computer from protocol address. Protocol address of next hop must be translated to h/w address before packet is sent

Address Resolution Translation from protocol address to hardware address known as Address Resolution. Address resolution is local to a network Computer can resolve address of another only if both are on the same physical network. Illustrated by the example below:

Address Resolution In fig, hosts A & B attach to same network If application on A sends data to B, the application uses B’s IP address as destination. Protocol s/w on A resolves B’s IP address to B’s h/w address and uses it to send frame directly If application on A sends message to application on F, software on A first determines that packet must travel through router R 1. Software on A resolves address of R 1 and

Address Resolution Software on R 1 determines that packet must reach R 2, resolves its address and sends packet Finally R 2 receives packet, determines that destination F is attached to rightmost network resolves address of F and delivers packet From example, each computer handling packet resolves next-hop address before sending

Address Resolution In summary, mapping between protocol and hardware address is address resolution Host/router uses address resolution to send packet to another computer on same network Computer never resolves address of host on a remote network

Address Resolution Techniques Algorithm used by s/w to translate protocol into hardware address depend on protocol and h/w addressing schemes. Example, method used to resolve an IP address to Ethernet address differs from that used to resolve it to an ATM address Since router can connect to multiple types of networks, it may use more than one type of address resolution Therefore, has more than one translation module

Address Resolution Techniques Address resolution algorithms grouped into three categories: 1. Table Lookup 2. Closed-form computation 3. Message exchange

Address Resolution Techniques 1. Table Lookup Bindings or mappings stored in a table in memory that software searches to resolve address 2. Closed-form computation Protocol address chosen carefully so that computer’s h/w address can be computed from protocol address using basic Boolean and arithmetic operation

Address Resolution Techniques 3. Message exchange Hosts exchange messages to resolve address One computer sends message requesting an address translation, another computer sends a reply containing requested information

Address Resolution with Table Lookup Requires data structure to contain information about address bindings Table consists of entries that contain protocol address and equivalent hardware address Fig below shows example of address binding table

Address Resolution with Table Lookup Each entry corresponds to one host on network Separate address binding table used for each n/w IP addresses in given table have same prefix Main advantage is generality: table can store address bindings for arbitrary set of computers on a network Protocol address map to arbitrary h/w address Table lookup algorithm for address resolution is straightforward and easy to program

Address Resolution with Closed. Form Computation Many networks technologies use static physical addresses, some use configurable addressing - network interface assigned a specific h/w address In such networks, possible to chose addresses that make closed-form address resolution possible Resolver that uses closed-form, computes mathematical function that maps an IP address to a hardware address. It is efficient for configurable networks because both hardware and IP addresses can be changed, thus values chosen to optimize translation

Address Resolution with Message Exchange Computer that needs to resolve address sends a message across network and receives a reply Message carries request specifying protocol address, and reply carries corresponding hardware address Address resolution request can be: ◦ sent to an address server which will then send a reply ◦ Broadcasted to all machines & corresponding machine responds

Address Resolution with Message Exchange Major advantage of first scheme is centralization Few address resolution servers handle all address resolution requests Address resolution easier to configure & manage Major advantage of second scheme is distributed computation Address resolution servers can be expensive and can become a bottleneck on a large busy network

Address Resolution Protocol (ARP) TCP/IP can use any of 3 resolution methods; Method chosen for network depends on addressing scheme used by underlying hardware Table lookup used for resolution across a WAN, closed-form computation with configurable networks, and message exchange on LAN hardware with static addressing

Address Resolution Protocol (ARP) To guarantee computers agree on format and meaning of messages used to resolve addresses, TCP/IP suite uses Address Resolution Protocol (ARP) ARP standard defines 2 basic message types: 1. request and 2. response Request message contains IP address and requests corresponding hardware address; Reply contains both the IP address sent in request and hardware address

IP Datagrams and Datagram Forwarding Goal of internet to provide system that allows application send data to another computer In well-designed internet, application program is unaware of: 1. 2. 3. underlying network to which it connects remote network to which destination connects interconnection between them Designers must decide whether to offer application connection-oriented service or connectionless service, or both

IP Datagrams and Datagram Forwarding TCP/IP designers chose to include both connectionless and connection-oriented services Chose to make fundamental delivery service connectionless, and add reliable connectionoriented service that uses underlying connectionless service Design successful, and often emulated by other protocols

Virtual Packets Connectionless internet service an extension of packet-switching Allows a sender to transmit individual packets of data across internet Each packet travels independently and has info identifying intended recipient Routers forward packets from one n/w to another

Virtual Packets Source host creates packet, places destination address in header, then sends to nearby router Router use destination address to select next router on path to destination then sends packet Eventually, packet reaches router that can deliver packet to final destination

IP Datagrams and Datagram Forwarding To overcome heterogeneity, IP software defines packet format independent of underlying h/w Result is universal virtual packet that can be transferred across underlying hardware Virtual implies protocol software handles packets in a way not understood by hardware Router connects heterogeneous networks; cannot transmit a copy of a frame that arrives on one network across another To accommodate heterogeneity, internet defines hardware-independent packet

The IP Datagram TCP/IP protocols use ‘IP datagram’ to refer to internet packets. Amount of data carried in datagram not fixed Sender chooses amount of data Size of datagram determined by sending application as appropriate for particular purpose Allowing size of datagrams vary makes IP adaptable to variety of applications Datagram can contain as little as 1 byte of data and at most 64 KB including header Header contains info to route datagram across internet e. g. it contains both sender and destination IP addresses

Forwarding an IP Datagram Each router along path to destination receives datagram, extracts destination address from header and uses it to determine next hop Router forwards datagram to the next hop: either final destination or another router To make selection of next hop efficient, each IP router keeps information in a routing table Routing table initialized when router boots and updated if topology changes or h/w fails Routing table contains set of entries each

Best-Effort Delivery IP defines semantic of communication, and use term best-effort to describe the service it offers The standard specifies that although IP makes best-effort attempt to deliver datagrams, does not guarantee that it will handle problems of: ◦ Data duplication ◦ Delayed or out-of-order delivery ◦ Corruption of data ◦ Datagram loss Additional layers of protocol software needed to handle each of these errors

IP Encapsulation, Fragmentation & Reassembly Datagram Transmission and Frames To forward datagram IP software in router/host selects next hop, N, then transmits datagram across physical network to N. Hardware doesn’t understand datagram format or internet addressing Each technology defines frame format and physical addressing scheme and will only accept packets that adhere to specified format and addressing scheme. Internet contains heterogeneous technology, frame format needed to cross a network may differ from frame format of previous network.

Encapsulation is used to transmit datagram across physical network that does not understand datagram format When IP datagram is encapsulated in a frame, entire datagram is placed in data area of frame Network hardware treats frame containing datagram exactly like any other frame.

Encapsulation To identify a frame whose data area contains an IP datagram sender and receiver agree on value used in frame type field. When sender places datagram in frame, it assigns frame type field a special value reserved for IP Frame arriving with special value in type field, receiver know data area contains IP datagram Encapsulation requires sender to supply physical address of next computer to receive datagram To compute IP address, software on

Encapsulation In summary, a datagram is encapsulated in a frame for transmission across a physical network. The destination address in frame is address of next hop to which the datagram should be sent; Address is obtained by translating IP address of the next hop to an equivalent hardware address.

Transmission across an Internet Encapsulation applies to a transmission at a time Sender selects next hop, encapsulates datagram in frame and transmits result across network When frame reaches, recipient removes IP datagram and discards frame If datagram must be forwarded to another network, new frame is created

Transmission across an Internet Fig below shows datagram encapsulated & un-encapsulated as it travels from source to destination through 3 networks and 2 routers Each n/w can use different hardware technology from others, meaning frame formats can differ.

Transmission across an Internet An IP datagram as it appears at each step during a trip across an internet. Whenever it travels across a physical network, the datagram is encapsulated in a frame appropriate to the network.

Transmission across an Internet As the figure shows, hosts & routers store datagram in memory with no additional header When datagram passes across network it is encapsulated in a frame suitable for the network Size of header on a datagram depends on network technology Example, if network 1 is an Ethernet, header in frame 1 is Ethernet header. Similarly, if network 2 is FDDI ring, header in frame 2 is FDDI header

Transmission across an Internet Note that frame headers do not accumulate during a trip through the internet. Datagram is encapsulated and frame header appended before it is transmitted across n/w When frame arrives at next hop, datagram is removed from frame before being routed and encapsulated in outgoing frame. When datagram reaches final destination, frame carrying datagram is discarded and it appears the same original size before being sent

MTU, Datagram Size and Encapsulation Each hardware technology specifies maximum amount of data that a frame can carry - Maximum Transmission Unit (MTU) Network hardware not designed to transfer frames that carry more data than MTU allows Datagram must be smaller or equal to n/w MTU or it cannot be encapsulated for transmission Problem in Internets with heterogeneous n/w s Router can connect networks with different MTU, it can receive datagram from n/w that

MTU, Datagram Size and Encapsulation Fig below illustrates router interconnecting two networks with MTU values of 1500 and 1000 w w w Host H 2 attaches to network with MTU of 1000 H 2 transmits datagrams of 1000 bytes or less H 1 attaches to network with MTU of 1500 bytes, can transmit datagrams that contain up to 1500 If H 1 sends 1500 datagram to H 2 , router R will receive datagram, but not be able to send it to w

MTU, Datagram Size and Encapsulation IP router uses technique known as fragmentation to solve problem of heterogeneous MTUs A datagram larger than MTU of network over which it must be sent is divided into smaller bits called fragments by router and sent independently Fragment has same format as other datagrams A bit in the flags field of the header identifies a fragment and full datagram. Other fields in the header are assigned information to assist reassemble the

MTU, Datagram Size and Encapsulation In summary, datagram cannot be larger than MTU of a network over which it is sent A router receiving a datagram larger than MTU of network over which it is sent, it divides it into smaller pieces called fragments Each fragment uses the IP datagram format, but carries only part of the data.

Reassembly Process of creating copy of original datagram from fragments is called reassembly Since each fragment has a copy of original datagram header all fragments have same destination address as original datagram Final fragment has an additional bit set in header Thus, a receiver performing reassembly can tell whether all fragments have arrived successfully IP specify that ultimate destination host reassemble fragments

Reassembly In fig, if H 1 sends a 1500 -byte datagram to host H 2, router R 1 will divide datagram into 2 fragments then forward to R 2 does not reassemble fragments, instead uses destination address to forward fragments Final destination host, H 2, collects fragments and reassembles them to produce the

Reassembly Advantages of destination reassembling include: 1. Reduces amount of state info in routers 2. It allows routes to change dynamically IP is free to pass fragments in different routes To reassemble fragments arriving out of order, sender laces unique id no in id field of datagram Router puts id no into each fragment when fragmenting Receiver uses id no and source IP address in fragment to determine datagram to which

Fragment Loss IP doesn’t guarantee datagram delivery – if network drop a packet, encapsulated datagram or fragment can be lost When all fragments arrive, datagram can be reassembled A problem arises when one or more fragments arrive, and some are delayed or lost

Fragment Loss Although datagram cannot be reassembled Receiver must save fragments in case missing fragments are only delayed Receiver cannot hold fragments indefinitely to avoid exhausting memory IP specifies a maximum time to hold fragments. When first fragment arrives, receiver starts timer If all fragments arrive before timer expires,

Fragment Loss If timer expires before all fragments arrive, receiver discards fragments that have arrived Result of IP’s reassembly timer is all-ornothing: All fragments arrive and datagram assembled or all discarded In poorly designed internet where networks are arranged in sequence of decreasing MTUs, each router along the path must further fragment each fragment

Fragment Loss IP doesn’t distinguish between fragments and sub-fragments Receiver cannot know whether fragment was result of a router fragmenting datagram or routers fragmenting fragments Advantage is that receiver can perform reassembly of original datagram without reassembling sub-fragments Reduces amount of info needed in headers & saves CPU time