Previous Table of Contents Next


6.5.5 FDDI

ANSI developed the Fiber Data Distributed Interface (FDDI) as a standard for fiber-optic data transmission. FDDI is a token-passing ring architecture that operates at 100 Mbps. (The actual data rate for FDDI is 125 Mbps, but 1 out of 5 bits handles overhead.) Because of its transmission rate, FDDI may emerge as a significant alternative to Ethernet or token ring for local data transport. The FDDI frame structure, shown in Figure 6-10, is similar to IEEE 802.6. The maximum frame size is 4500 octets (or 9000 symbols, with 4 bits/symbol). When 6-octet addressing (the most common) is used, the MAC-layer header (preamble through source address) uses 16 octets, and the MAC-layer trailer uses 6 octets. Subtracting the headers from the maximum frame size leaves 4478 octets for data. The IEEE 802.2 LLC header requires three octets, yielding a maximum IP datagram length of 4475 octets. (Because FDDI does not typically use SNAP, the figure does not show the SNAP header, but there is no technical constraint on the use of SNAP. If present, the SNAP header would immediately follow the IEEE 802.2 header.) RFC 1188 [6-9] defines support for FDDI within TCP/IP-based internetworks.


Figure 6.10.  FDDI frame with SNMP message (Courtesy American National Standards Institute)

6.6 Address Translation

In the previous two sections, we discussed the differences between the Internet address, used by IP, and the local address, used by the LAN or WAN hardware. We also saw that the IP address is a 32-bit logical address, but the physical hardware address depends on the hardware. For example, ARCNET has an 8-bit hardware address and Ethernet has a 48-bit hardware address. Thus, translation between the physical and logical addresses is necessary. The Address Resolution Protocol (ARP) described in RFC 826 [6-10] translates from an IP address to a hardware address. The Reverse Address Resolution Protocol (RARP), detailed in RFC 903 [6-11], does the opposite, as its name implies.

6.6.1 Address Resolution Protocol (ARP)

Assume that a device on an Ethernet, Host X, wishes to deliver a datagram to another device on the same Ethernet, Host Y. Host X knows Host Y’s destination protocol (IP) address, but does not know Host Y’s hardware (Ethernet) address. Host X would therefore broadcast an ARP packet (shown in Figure 6-11) on the Ethernet to determine Host Y’s hardware address. The packet consists of 28 octets, primarily addresses, contained within the Data field of a local network frame. A device that recognizes its own protocol address responds with the requested hardware address. The individual fields of the ARP message show how the protocol operates.


Figure 6-11.  Address Resolution Protocol (ARP) and Reverse Address Resolution Protocol (RARP) packet formats

The first field, Hardware (2 octets), defines the type of hardware in use. Current values are listed in RFC 1700. Examples include Hardware = 1 (Ethernet), 6 (IEEE 802 Networks), 7 (ARCNET), and 11 (LocalTalk). The second field, Protocol (2 octets), identifies the protocol address in use. For example, Protocol = 0800H would identify IP addresses.

The next two fields allow the ARP packet to work with addressing schemes of different lengths (Figure 6-11 represents the most common scheme, where 6 octets are required for the Hardware Address and 4 octets are required for the Protocol Address). To make the protocol more adaptive, the HA Length (1 octet) and PA Length (1 octet) specify the lengths, in octets, of the addresses to be used. Figure 6-11 represents the most common scenario. Six octets (48 bits) are required for Hardware Address (HA Length = 6) and 4 octets (32 bits) for the Protocol Address (PA Length =4 ). The Operation field (2 octets) defines an ARP Request = 1 or ARP Reply =2.

The next fields contain the addresses themselves. With an ARP Request message, the Target Hardware Address (HA) field is unknown and is sent filled with zeros. The ARP Reply packet from the target host inserts the requested address in the field. When it receives the ARP Reply, the originating station records the information in a table (known as the ARP cache) so that it doesn’t need to make the same request again and again. Routers have an ARP cache with a finite lifetime so the table won’t grow too large.

6.6.2 Reverse Address Resolution Protocol (RARP)

Most network hosts are intelligent enough to remember their hardware and protocol addresses. Diskless workstations, however, depend on a server to provide much of their intelligence. The diskless workstation would know its hardware address (HA), which is coded into its ROM, but it may not know its protocol address (PA), which the server assigns. The RARP can discover the unknown PA given a known HA and a RARP server to supply the desired answer.

The process of determining an unknown protocol address is similar to that of finding an unknown hardware address. The same packet structure is used (review Figure 6-11), with only minor modifications to the field values required. The Operation field adds two new values, 3 (RARP Request) and 4 (RARP Reply). When the RARP Request is made, the Sender HA, Sender PA, and Target HA are transmitted. The RARP Reply contains the requested Target PA.


Previous Table of Contents Next