arp rarp

Introduction

• IP addresses are “virtual”, maintained by SW• A frame transmitted across a physical NW

– must have the HW address (HWA) of the destination

• SW must translate the protocol address (PRA) of the destination into an equivalent HWA

This section • describes three general mechanisms that are used to

perform address mapping:– 1st relies on a table that contains address translation information– 2nd uses a mathematical function to perform the translation.– 3rd makes two computers exchange messages across a NW

Protocol Addresses And Packet Delivery

• SW in each host or router uses the destination PRA to select a next hop for the packet

• PRA cannot be used when transmitting frames across physical NW– the HW does not understand PRA

• A frame must use the HW's frame format– and all addresses in the frame must be a HWA

• PRA of a next hop must be translated to an equivalent HWA before a frame can be sent

Address Resolution (AR)

• Mapping between a PRA and a HWA – called “address resolution” (AR)

• A host/router uses AR when it needs to send a packet to another computer on the same physical NW– AR is local to a NW

• A computer never resolves the address of a computer that attaches to a remote NW.

• Each computer that handles a packet resolves a next-hop address before sending

Address Resolution Techniques

• Different methods needed– To resolve a PRA to an Ethernet address differs from the

method used to resolve a PRA to an ATM address

• AR algorithms can be grouped into three categories:• Table lookup

– Bindings or mappings are stored in a table in memory

• Closed-form computation– HWA can be computed from the protocol address using basic

Boolean and arithmetic operations

• Message exchange– Computers exchange messages across a NW for AR; “request” and “reply” messages

AR With Table Lookup

• Table is an array with a pair of ( P, H ) entries – P PRA– H equivalent HWA

• A separate binding table is used for each NW• Table in next Figure corresponds to a NW with prefix

197.15.3.0 / 24. Each IP address has 197.15.3 prefix– We can save space by omitting the prefix from table entries

• Chief advantage of the table lookup is generality– a table can store the address bindings for an arbitrary set– arbitrary no. of PRAs can map to an arbitrary no. of HWA

AR With Table Lookup (cont)

• Sequential search for a match for a given PRA• For large NWs a sequential search requires excessive

CPU time– For computational efficiency– We can use “hashing” or “direct indexing”

• Administrators may choose non-sequential numbers for PRA to help identify the purpose of a computer, e.g. – hosts with suffixes < 200 , while routers with suffixes > 200– host suffix from an IP address as an index into the array

• To prevent an illegal PRA from causing a subscript error– must check to ensure the suffix is in range

AR With Closed-Form Computation

• Some technologies use configurable addressing– A NW interface can be assigned a specific HWA– values can be chosen to optimize the translation– especially, the host portion of PRA can be chosen to be identical

to the computer's HWA

• HWA can be computed by a Boolean and (&) operation: HWA = PRA & 0xFF

AR With Message Exchange

• A request HWA corresponding to a PRA• Reply carries the corresponding HWA• Where should an AR request be sent?• Most protocols choose one of two possible designs

– A NW may include one or more servers • contact servers in turns

• broadcasts its request to all servers

– Each computer on the NW participates by agreeing to answer resolution requests for its address:

• broadcasts a request on the NW, all machines examine

• If request matches a computer's address, then responds

• Centralization versus distributed mechanism ?

Address Resolution Protocol

• TCP/IP can use any of the three methods depending on the addressing scheme used by HW– Table lookup to resolve PRA across a WAN– Closed-form computation for configurable NWs – Message exchange on LAN with static/dynamic addressing

• TCP/IP has an Address Resolution Protocol (ARP), which defines two message types: – A request for a HWA correspond to an PRA (namely IP address)– A reply returning the HWA

ARP Message Delivery

• An ARP request message should – Be placed in a HW frame and broadcast to all computers

• All receive the request and examines the IP address– The computer mentioned in the request sends a response; – All others discard the request

• The ARP response is not broadcast– Response is sent directly back to the computer that issued the

request

ARP Message Format

• ARP describes the general form for messages, – and specifies how to determine the details for each type of HW

• To increase the generality of ARP– the designers included an address length field for PRA and HWA

• ARP is not restricted to IP addresses or specific HWA– In practice, these generality of ARP is seldom used:– ARP is almost always used to bind a 32-bit IP address (a PRA)

to a 48-bit Ethernet address (a HWA)

Sending An ARP Message

• When sending an ARP, the message travels inside a HW frame

• The ARP message is treated as data being transported – the NW HW does not know about the ARP message format and

does not examine the contents of individual fields

• Placing a message inside a frame for transport is called encapsulation– an ARP message is encapsulated directly in a HW frame

• Figure 19.7 illustrates the concept

Identifying ARP Frames

• How does a host know whether an incoming frame contains an ARP message?– The type field specifies that the frame contains an ARP msg

• Sender must assign the appropriate value to the type field before transmitting the frame

• Receiver examine the type field in each incoming frame

• Figure 19.8 illustrates the concept

Processing An Incoming ARP Message

• Receiver perform two steps when a message arrives – extracts the sender's address binding– checks to see if the sender's address is present in the cache

• ARP contains a further optimization:– after a reply to an ARP request, adds to its cache for later use

• Most computer communication involves two-way traffic – if a message travels from one computer to another, probability is

high that a reply will travel back

• Because memory constraints– only the target of an ARP request perform caching