ARP Caching
ARP is a dynamic resolution protocol, which means that every resolution requires the interchange of messages on the network. Each time a device sends an ARP message, it ties up the local network, consuming network bandwidth that cannot be used for ?real? data traffic. ARP messages aren't large, but having to send them for every hop of every IP datagram would represent an unacceptable performance hit on the network. It also wastes time compared to the simpler direct mapping method of resolution. On top of this, the ARP Request message is broadcasted, which means every device on the local network must spend CPU time examining the contents of each one.
The general solution to the efficiency issues with dynamic resolution is to employ caching, which I described in general terms in the section on address resolution concepts. In addition to reducing network traffic, caching also ensures that the resolution of commonly-used addresses is fast, making overall performance comparable to direct mapping. For this reason, caching functionality has been built into ARP from the start.
Static and Dynamic ARP Cache Entries
The ARP cache takes the form of a table containing matched sets of hardware and IP addresses. Each device on the network manages its own ARP cache table. There are two different ways that cache entries can be put into the ARP cache:
* Static ARP Cache Entries: These are address resolutions that are manually added to the cache table for a device and are kept in the cache on a permanent basis. Static entries are typically managed using a tool such as the arp software utility.
* Dynamic ARP Cache Entries: These are hardware/IP address pairs that are added to the cache by the software itself as a result of successfully-completed past ARP resolutions. They are kept in the cache only for a period of time and are then removed.
A device's ARP cache can contain both static and dynamic entries, each of which has advantages and disadvantages. However, dynamic entries are used most often because they are automatic and don't require administrator intervention.
Static ARP entries are best used for devices that a given device has to communicate with on a regular basis. For example, a workstation might have a static ARP entry for its local router and file server. Since the entry is static it is always found in step #1 of the ARP transaction process, there is no need to ever send resolution messages for the destination in that entry. The disadvantage is that these entries must be manually added, and they must also be changed if the hardware or IP addresses of any of the hardware in the entries change. Also, each static entry takes space in the ARP cache, so you don't want to ?overuse? static entries. It wouldn't be a good idea to have static entries for every device on the network, for example.
Proxy ARP
ARP was designed to be used by devices that are directly connected on a local network. Each device on the network should be capable of sending both unicast and broadcast transmissions directly to each other one. Normally, if device A and device B are separated by a router, they would not be considered local to each other. Device A would not send directly to B or vice-versa; they would send to the router instead at layer two, and would be considered ?two hops apart? at layer three.
Why Proxy ARP Is Needed
In contrast to the normal situation, in some networks there might be two physical network segments connected by a router that are in the same IP network or subnetwork. In other words, device A and device B might be on different networks at the data link layer level, but on the same IP network or subnet. When this happens, A and B will each think the other is on the local network when they look to send IP datagrams.
In this situation, suppose that A wants to send a datagram to B. It doesn't have B's hardware address in the cache, so it begins an address resolution. When it broadcasts the ARP Request message to get B's hardware address, however, it will quickly run into a problem: B is in fact not on A's local network. The router between them will not pass A's broadcast onto B's part of the network, because routers don't pass hardware-layer broadcasts. B will never get the request and thus A will not get a reply containing B?s hardware address.
More info here:
http://www.tcpipguide.com/free/t_ProxyARP-2.htm
Please post back if you have any more problems or questions.