Previous | Table of Contents | Next |
As we studied previously, the IEEE 802.3 standard, which is covered by the RFC 1042 [6-7] Internet standard, is similar but not identical to the DIX Ethernet. Figure 6-7 shows the IEEE 802.3 format. IEEE 802.3s Destination and Source address fields may be 2 or 6 octets long, although the 6-octet length, which matches the Ethernet address lengths, is most common. The Address Resolution Protocol (ARP) maps the IP address (32 bits) to the IEEE 802 address (48 bits). The ARP hardware code for IEEE 802 networks is 6. However, broadcast addresses for both Ethernet and IEEE 802 networks are consistent with all Ones.
Figure 6-7. IEEE 802.3 frame with SNMP message (©1990, IEEE)
Next, the IEEE 802.3 frame defines a Length field, which specifies the length of the Data unit. Recall that in the Ethernet frame this was the Ethertype, indicating the higher-layer protocol in use. These two octets (the Ethertype or Length fields) distinguish whether the frame format is Ethernet or IEEE 802.3, respectively. If the Data Link layer driver mixes these up, confusion will result. For example, if the destination host expects an Ethernet frame (with an Ethertype field), it cannot respond to an IEEE 802.3 frame containing the Length field.
The Data field contains the information from the higher layers, plus two IEEE-defined headers. The first header is the Logical Link Control (LLC) header defined by IEEE 802.2. The LLC header includes destination and source service access point addresses (DSAP and SSAP, respectively) and a control field. The second header is known as the Sub-Network Access Protocol (SNAP), defined by IEEE Standard 802-1990. This header includes a Protocol ID or Organization Code field (3 octets) and an Ethertype field (2 octets). The combination of the LLC and SNAP headers allows the higher-layer protocol in use to have both an SAP and Ethertype designation. The rest of the Data field contains the higher-layer information, such as an IP datagram.
The IEEE 802.5 token ring has enjoyed great success, partly because of strong support from major networking companies such as Apple, IBM, and Proteon, and partly because of the protocols built-in provision for internetworking. This provision is known as source routing and uses the Routing Information (RI) field to connect rings via bridges. The RI field specifies the path the frame must take from its source to its destination. The mechanism for determining that path is called route discovery.
The IP Datagram occupies the Information field of the token ring frame, as shown in Figure 6-8. Any necessary Routing Information precedes the Information field. Inside the Information field is the IEEE 802.2 LLC header (3 octets), SNAP header (5 octets), and the IP datagram (variable length). Given a minimum IP datagram header of 20 octets, the protocol overhead (LLC + SNAP + IP) is 28 octets per IP datagram. The maximum length of the Information field (and thus the encapsulated IP datagram) varies, depending on the token holding time parameter. This parameter specifies the length of time a particular node may hold the token before it must pass the token to its downstream neighbor. RFC 1042 [6-7] discusses an example token holding time of 9 milliseconds, which results in a maximum length of the IP header plus datagram of 4464 octets.
Figure 6-8. IEEE 802.5 frame with SNMP message
Datapoint Corporation developed the Attached Resource Computer Network (ARCNET) in 1977. ARCNET is a token-passing architecture that can have several Physical layer implementations, including a linear bus, a star, or a branching tree. The original version supported a transmission rate of 2.5 Mbps and up to 255 workstations, and is standardized as ANSI 878.1. In February 1992, Datapoint announced ARCNETPLUS, a 20 Mbps network which is downward-compatible with the original ARCNET.
The Internet standard for ARCNET, RFC 1201 [6-8], suggests methods for encapsulating both IP and ARP datagrams within the ARCNET frame. Three frame formats are available, as shown in Figures 6-9a, b, and c. Note that this RFC supersedes the older version (RFC 1051) and makes a number of protocol enhancements for improved TCP/IP support. The short frame format (Figure 6-9a) limits the transmitted client data to 249 octets. The long frame (Figure 6-9b) allows between 253 and 504 octets of client data. An exception frame (Figure 6-9c) is used with frames having 250 to 252 octets of client data. (Note that the frame formats shown in Figure 6-9 appear in the software buffers. The format transmitted by the hardware duplicates the Destination ID [DID], does not send the Unused and Protocol ID fields, and adds some hardware framing.)
Figure 6-9a. ARCNET short frame format with SNMP message
Figure 6-9b. ARCNET long frame format with SNMP message
Figure 6-9c. ARCNET exception frame format with SNMP message
The ARCNET frame may contain up to 512 octets, of which 504 may be client data. The sender fragments larger packets, using the Split Flag and Sequence Number fields to identify the fragments. The Split Flag takes on one of three values depending on the fragmentation required. Unfragmented packets use Split Flag = 0. The first fragment of a fragmented packet uses Split Flag = ((T-2)*2)+1, where T = total number of expected fragments. Subsequent fragments use Split Flag = ((N-1)*2), where N is the number of this fragment.
For example, assume that a packet requires 8 fragments. The Split Flag values would be:
Fragment | Split Flag (decimal) |
---|---|
1 | 13 |
2 | 2 |
3 | 4 |
4 | 6 |
5 | 8 |
6 | 10 |
7 | 12 |
8 | 14 |
Up to 120 fragments are allowed, yielding a maximum value of 238 decimal (EEH). This allows up to 60,480 octets per packet (120 * 504 = 60,480). All fragments belonging to the same packet use an identical 2-octet Sequence Number.
Datapoints ARCNETPLUS standard permits a Data field of 4224 octets in length. Currently, no Internet standard (RFC) exists that discusses the fragmentation requirements for this network.
Another unique characteristic of ARCNET and ARCNETPLUS is their addressing structures, which define an 8-bit address field and allow 255 unique hardware addresses as well as a broadcast designation (address = 0). (This address is implemented with an 8-position DIP switch, which you must set manually on each ARCNET or ARCNETPLUS card. A human error in duplicating these switch settings can cause ARCNET to fail; ARCNETPLUS offers an algorithm to detect duplicate addresses.)
Previous | Table of Contents | Next |