Overview of ARCnet

ARCnet stands for Attached Resource Computer Network. Because it is a token-passing system, ARCnet is a deterministic network technology that is useful in situations where a predictable throughput is required. It was a very popular technology during the early 1980s, when Ethernet was still quite expensive and most local area networks were small. Its speed of 2.5Mbps was more than sufficient for implementing ARCnet in a small office network, given the relative power of PCs and minicomputers at that time. Now, it is most likely to be found in older departmental LANs (though I've never seen one) or in an industrial manufacturing plant or another similar setting. The basic ARCnet operates at a rate of 2.5Mbps and can be used to create a LAN of as many as 255 computers. Some network hardware vendors produce network adapters and hubs that allow for speeds up to 10Mbps.

Although ARCnet is no longer marketed primarily as a PC LAN solution, it does have many features that make it well suited for industrial applications. Factory floor automation requires that controllers and other devices have a communications network in place that allows for reliable, predictable throughput. The following are some of the reasons why ARCnet is still in use today, in environments such as this:

• It's deterministic. That is, because it uses a token-passing mechanism, it is possible to calculate the worst-case amount of time it takes to get a frame from one node to another. Using basic Ethernet, nodes on the network must contend for access to the shared network medium, and performance can suffer as network traffic increases .

• It's simplistic. Other than assigning an address to each ARCnet node, few software configuration or management tasks are needed. This makes ARCnet practically "invisible" to workers using devices that have embedded ARCnet adapters.

• Although ARCnet is similar to Token-Ring technology in that it uses a token frame to grant access to the network, no central computer or device is responsible for monitoring or managing the network. All nodes in the network are equal peers. Adding and removing devices from the network is a simple task.

• ARCnet can be wired using various cablesfrom coaxial to fiber optic . This makes connecting devices or adapters from multiple vendors easy.

• Although adapter cards for use in PCs and other computer are generally more expensive than Ethernet alternatives, chips used in embedded controllers are usually quite inexpensive.

In an environment such as factory automation, the fact that ARCnet requires little configuration and management makes it a good solution.

Although development was started before the OSI reference model was defined, ARCnet provides functions along the same lines as those defined in the physical and data link layers of the reference model. The ARCnet network adapter card, or the embedded chip in a factory device, takes care of the successful, reliable transmission of a message, relieving the software protocol of these functions.

See "Overview of the OSI Seven-Layer Networking Reference Model," p. 1121.

Although Datapoint originally manufactured the chips used to create ARCnet network adapters, the primary manufacturer of chips used today is Standard Microsystems Corporation (SMSC). You can reach their Web site at www.smsc.com. This site contains specific technical documentation for the chips found in most ARCnet products today.

ARCnet Addressing and Message Transmission

The logical topology of the ARCnet network is always a token bus, although it can be physically arranged as a bus, a star, or a hierarchical star topology, which is a combination of the two (see "Bus and Star Topologies," later in this chapter). ARCnet is a logical token bus because no matter which physical topology is used, the token frame , which grants permission to a node to transmit data, is passed around in a sequential manner based on a numerical address from one node to the next . Thus, all nodes get an equal chance to access the network media within a maximum set time limit.

See Chapter 2, "Overview of Network Topologies," p. 13.

See the chapter "Token-Ring Networks," located on the upgradingandrepairingpcs.com Web site.

Note

ARCnet isn't restricted to the factory floor. For Linux users there is even a Linux How-To document, maintained by Avery Pennarun, that can guide you in choosing a card and a network driver so that you can link your Linux computers using ARCnet. You can find this How-To at www.worldvisions.ca/~apenwarr/arcnet/howto/index.html.

Each node on the LAN is configured with an address from within the range of 1255. This is because the address fields in ARCnet frames are only 8 bits in length (one byte). Because the largest number you can store in a single byte is 255, this limits ARCnet LANs to small implementations . Of course, back in the 1970s, linking 255 computers was considered exceptional.

The token frame, called an Invitation to Transmit (ITT), is sequentially passed from one node to another based on the addresses assigned to devices in the network. When a node receives the ITT token, it can then transmit a message on the network, or it can pass the frame to the node that has the next highest numerical address in the network. The next node doesn't have to be the node that is located physically closest to the sending node; it can be anywhere on the network. The hierarchy of numerical addressing determines the order in which a node is granted permission to transmit on the network.

ARCnet Symbols and Frame Formats

Two different frame formats are used on an ARCnet LAN. The basic frame format consists of five types, each of which is used for a specific messaging purpose. The second frame format is the Reconfiguration Burst frame, which is the only frame of the five that is used in configuring the network. All frames, however, are made up using a set of basic symbols (see Figure 13.1):

• Start Delimiter (SD) This symbol consists of six ones. All the basic frames use this symbol to indicate the beginning of a frame.

• Reconfiguration Symbol Unit (RSU) This symbol is made up of eight ones followed by a zero. This symbol is used in a special frame, described later in this chapter. It is used to reconfigure the network when a node joins the network.

Figure 13.1. Frame types used on ARCnet LANs.

Besides these two symbols, a set of Information Symbol Units (ISU) is used. Each ISU consists of a bit pattern of 110 followed by an 8-bit value for each ISU. Thus, all ISUs are 11 bits in length. ISUs can serve to indicate the kind of frame that is being transmitted, network addresses, and the actual data that is transmitted inside a data packet frame. Most of the frames you will see on an ARCnet LAN consist of the SD symbol followed by one or more of the following ISUs:

• Start of Header (SOH) This value indicates the beginning of a data packet. Note that the SOH should not be confused with the SD symbol, which precedes the SOH symbol. The SD symbol indicates the start of the frame, whereas the SOH symbol indicates that the frame contains a data packet. The value for this symbol is 0x01.

• Enquiry (ENQ) This symbol is used in a frame that is sent to determine whether the destination node has enough buffer space in memory to receive a message. The value for this symbol is 0x85.

• Acknowledgment (ACK) This symbol is used to send an acknowledgment. The value for this symbol is 0x86.

• Negative Acknowledgment (NAK) This symbol is used as the opposite of an ACK. It is used in a frame to tell the sender that the destination node does not have free buffer space at this time. The value for this symbol is 0x15.

• End of Transmission (EOT) This symbol is used in the token frame. The node that receives a frame containing this symbol can then begin to transmit a message on the network medium. The value for this symbol is 0x04.

• Next Node Identification (NID) This symbol contains the address of the next logical node in the network and also is used in the token frame. This address is the node to which the token frame will be passed when the current holder releases it. The values of this symbol can range from 0x01 to 0xFF (1255 decimal).

• Source Node Identification (SID) This symbol contains the address of the sender of a data message packet. Like the NID, this symbol can have a value ranging from 0x01 to 0xFF.

• Destination Node Identification (DID) This symbol contains the address of the node to which a request to send frame or a data packet frame is being sent. Like the NID and SID, this value can range from 0x01 to 0xFF.

• Continuation Pointer (CP) This symbol indicates the length of the data packet. The value of this symbol can range from 0x03 to 0xFF. As explained later in this chapter, data packets come in two types: short and long. Short packets contain one CP symbol, whereas long packets contain two CP symbols.

• System Code (SC) This symbol is assigned by the ARCnet Trade Association, mentioned earlier in this chapter, and is used to indicate a higher-level protocol. SC symbols are used in data packets. The value for this symbol can range from 0x00 to 0xFF. Using a symbol to specify the protocol allows more than one network protocol to be used on the same network. Note that the value of 0x80 is a reserved code and is used for diagnostic purposes only.

• Frame Check Sequence (FCS) This symbol contains the value of the calculated cyclic redundancy check (CRC-16) used to verify that the contents of the data packet arrived intact and were not corrupted during transit. The value for this symbol can range from 0x00 to 0xFFFF. Because the value of 0xFFFF requires more than 8 bits in binary, two ISU symbols are used to create the FCS in a data packet.

• Data The actual user data contained in a packet is composed of multiple ISUs, each 11 bits in length. Although you need only 8 bits to create a byte, remember that each ISU contains the 3-bit preamble of 110, so each byte of user data is actually represented by 11 bits inside the data portion of a data packet.

Using these basic symbols, it is now possible to define the kinds of frames that can be constructed so that nodes can communicate on the ARCnet LAN. The five basic frame types are listed here:

• ITT Invitation to Transmit. This is the token frame that is passed around the logical network, giving the node that possesses the frame permission to send a data message on the network medium.

• FBE Free Buffer Enquiry. This frame type is used by a node to determine whether the destination node has sufficient buffer space in memory to receive a message before it is sent.

• ACK Acknowledgment. This frame is sent in response to the FBE frame if the destination is willing to receive the message.

• NAK Negative Acknowledgment. This frame is sent in response to the FBE if the destination is not willing to receive the message at this time.

• PAC Packet. This frame carries the actual message data. ARCnet uses two kinds of PAC frames: short and long. Obviously, the long version is meant for sending messages of a larger size.

Figure 13.1 shows the layout of the symbols used to construct the different frame types.

Note that in Figure 13.1, each frame type begins with the start delimiter (SD) symbol. Only the actual data packet frames (PAC) contain the frame check sequence that is used to verify the integrity of the data in the packet. Another interesting thing to note is that a short packet can contain as many as 2,772 bits in the data section, whereas the long packet type has a minimum of 2,816 bits for the data section. This is because data ISUs with values of 253, 254, or 255 are not allowed. If a message falls within these bounds, the message is sent in a long packet, with null padding (a string of zero bytes) added to adjust the length of the packet.

Note also that the number of bits includes the 3-bit preamble for each byte of data in the data portion of the packet. That is, although the data section in a short packet can be up to 2,772 bits long, that doesn't mean you can place 346.5 (2,772/8) bytes of actual data in the data section. Instead, you can place only a maximum of 252 (2,772/11) bytes of data. Each byte effectively is represented by 11 bits, with the first 3 bits being the constant value of 110. Likewise, the maximum number of bits in the data portion of a long data packet is 5,577, which means that the maximum number of bytes that a long packet can carry is 507 bytes (5,577/11).

The following basic steps are involved in sending a message in an ARCnet network:

1. A node that has a message to send receives the token frame.

2. Before sending the message, the source node sends a Free Buffer Enquiry (FBE) frame to the node to which it wants to communicate. Looking back at Figure 13.1, you can see that this is a simple frame that doesn't even contain the source address of the computer or device making the request. Only the destination address is contained in the frame, and it is stored in the frame twice.

3. If the destination node has available buffer space to receive a message, it transmits an acknowledgment frame. If it does not have sufficient space in its memory to receive the message, it sends a negative acknowledgment. Again, this frame does not contain the source address of the device that originated the request.

4. When the sender sees the acknowledgment frame, it sends the message to the destination node. The message is sent in a packet that can range up to as many as 507 bytes of data. As with most network technologies, the message packet contains a header. Both the source and the destination addresses are included in this frame. This frame uses a CRC (cyclic redundancy check) value that the receiving node can use to ensure the accuracy of the message received.

5. When the receiving node checks the CRC value against the data and determines that the message was accurately received, it sends another acknowledgment frame around the network. If the message data fails the CRC test, the receiving node transmits nothing. A timer on the sending node eventually expires , so the sender knows that its message was not delivered correctly.

6. After a successful transmission (acknowledgment received) or unsuccessful transmission (timer expired ), the sending node relinquishes the token by passing it to the network node that has the next higher address on the network.

In this sequence of events, the node that gains control of the token frame can make only one attempt at transmitting a data packet. If the data does not reach its intended destinationthe timer expires before an acknowledgment is receivedthe sending node does not immediately try to resend the data packet. Instead, it passes the token frame (ITT) to the next logical member of the ring and waits until the token frame returns to it again before it attempts to send the data packet.

If the message that the sender wants to transmit is larger than 507 bytes, the sender divides the message into smaller units and must wait until it receives the token again to send each fragment. The receiving node reassembles the fragments to get the full message.

Note

ARCnet also supports a broadcast message, using zero as the destination address. When this type of message is sent, the acknowledgment procedure is not used. Additionally, only nodes that have been configured to respond to broadcast messages will process them.

This simple scheme shows that it is possible to calculate the minimum amount of time it can take to send a message from one node to another. It also shows that the worst-case scenario can be calculated by taking into account the number of nodes on the network, cable lengths, timer values, and other similar factors. Thus, for real-time applications in which network transmit time needs to be predictable, ARCnet can be a good solution. For example, the more nodes you add to the network, the greater the latency time before a node will be able to transmit. But it will, with some exceptions, be able to transmit within a specific time, which can increase as the number of connected nodes increases. This does not take into account the fact that not all nodes are waiting to transmit data. That depends on the type of network and the devices attached.

Network Configuration

ARCnet does not require that the network administrator assign addresses sequentially beginning with 1 and continuing through 255. In fact, as long as each node has a unique address assigned to it, it doesn't matter whether there are gaps in the address space. The ARCnet LAN uses a process that automatically lets each node discover its logical neighbor (the node with the next highest numerical address). When adding or removing nodes from the network, the network undergoes an automatic reconfiguration process.

The method by which a node joins the network uses the second frame format used by ARCnet: the Reconfiguration Burst frame format. This frame contains 765 RSU symbols (eight ones followed by a zero). When a node determines that it is not part of the logical LAN (that is, it doesn't receive the token within a short period) it uses the Reconfiguration Burst frame to effectively disrupt any current transmission or token passing that is in progress at the time. Other nodes on the network then begin a process of reconfiguration.

Each node backs off for a timeout period based on its numerical address. The node that has the highest address will be the first node to timeout. It then will attempt to locate its logical neighbor by incrementing its own address by one and transmitting a token frame. If a node with that address does not respond, the node increments the address again and continues to send a token frame until its neighbor is found. The remaining nodes on the network use this process so that, within a very short period, each node in the network knows its neighbor and normal communications can resume.

A node leaving the network is detected easily, and the network again undergoes a reconfiguration process. After a node sends a token frame to its neighbor, it continues to listen on the network to be sure that the neighbor will in turn pass on the token, or perhaps start the process of sending a message to another node. Remember that, although the ARCnet LAN can be laid out in bus and star formations, the address space (1255) makes the network a logical ring so that a node can be sure that it will hear something on the network within certain time limits. In the original standard, the maximum time allowed for a frame to make the trip from one node to another is 32 microseconds. In actual applications, it is possible to extend this value by adjusting timers to allow for greater distances between nodes.

Thus, when a node determines that something has gone wrong with its current logical neighbor, it then starts to search for a new neighbor. It does this by taking the address of the neighbor that was removed from the network (or failed in some way) and incrementing it by one. It sends out a token based on this new address. This continues until some response is detected (that is, the token successfully passes around the net).

Hubs and Network Wiring

An ARCnet network can be wired using universal twisted-pair (UTP) cables, coaxial cables, or fiber-optic cables. For UTP, Category 3 cables or above should be used. Coaxial cables should be RG11U or RG-59U or RG-62. The distance between network nodes depends on the type of wiring used and the type of hubs used as wiring concentrators .

Two types of hubs can be used when creating an ARCnet network:

• Active hubs Active hubs provide the longest distance capabilities for the ARCnet LAN. The active hub acts much like an Ethernet hub, and it is usually manufactured in 8- to 16-port units. The active hub takes the incoming signal and amplifies it before sending it back out on the other ports. Some active hubs perform other tasks, such as segmenting or blocking off a port that exhibits errors so that other segments are unaffected.

• Passive hubs Passive hubs usually have only four ports and do no signal amplification. Instead, a passive hub acts as a simple signal-splitter, taking the incoming signal and dividing it among the other three ports. A passive hub can be used to create a very small LANthat is, one with four nodes or fewer. The primary function of a passive hub in a larger LAN is to join individual workstations to an active hub. Unused ports on a passive hub usually should be terminated. Unused ports on active hubs, depending on the manufacturer, might or might not need to be terminated .

Bus and Star Topologies

Even though ARCnet is a token-passing technology, like a Token-Ring network, it can be wired in using several methods . Unlike with Token-Ring networks, no computer on the ARCnet LAN acts as a monitor to check for errors or otherwise manage the network. The reconfiguration process that ARCnet uses to form a logical ring is not directly related to the physical layout of the network.

Bus Topology

The simplest network topology that can be used for ARCnet is a bus using coaxial cables with BNC T-connectors. Up to eight nodes can be connected to any bus segment in a daisy-chain made up using the T-connectors. The total length of the segment is limited to 300 meters (1,000 feet). For such a small network, no hub would be necessary. However, as you can see in Figure 13.2, an active hub can be used to join multiple segments to create a larger LAN than can be created by using a single cable segment. A passive hub cannot be used to connect individual segments based on a bus topology; only active hubs can do this. Passive hubs can be used only to connect individual workstations.

Figure 13.2. An active hub can be used to join multiple coaxial segments, and might need terminators.

You also can create a bus topology using UTP cables. The UTP ARCnet adapter has two connectors, usually RJ11 or RJ45. Stations are daisy-chained from one node to the next using both connectors. In some cases, the last node on each end of the bus will need to have a terminator inserted into the last connector. Some cards provide an auto-termination feature. When using UTP, you can have as many as 10 nodes on one segment, with any repeater counting toward that limit. Each node on the bus must be separated by a minimum of about 6 feet. The total segment length can be as much as 400 feet (120 meters).

Star Topology

You can create a physical star topology by using hubs. You can create a tree structure of multiple stars by cascading hubs. Figure 13.3 shows a small network that uses both active and passive hubs, and also has workstations that connect directly to an active hub.

Figure 13.3. The star topology can be extended using additional hubs.

The major difference you will notice in Figure 13.3 is that, when a station connects to an active hub, the distance can be as great as 2,000 feet (500 meters). When a station connects to a passive hub, the distance shrinks to only 100 feet (30 meters). These same rules apply when connecting hubs. The passive hub connected to an active hub cannot be farther away than 100 feet. Two active hubs can be separated by as much as 2,000 feet. The capability of the active hub to regenerate the signal accounts for the longer distances achieved. The overall size of the network should never exceed 20,000 feet (6,000 meters).

Note

You can use active hubs to reach the maximum distance of 6,000 meters. When a large LAN is constructed in this manner, remember that no loops can exist in the physical topology and the total number of network nodes is limited to 255 because of the 8-bit address used.

Many vendors sell hubs, cables, and other network devices that can be used to create ARCnet networks that vary from these topologies and their limitations. The best source for information about vendors who supply ARCnet hardware is the ARCnet Trade Association (www.arcnet.com).

ARCnet Network Adapter Cards

Because ARCnet has been around for so long, a lot of different network cards are still in use today. ARCnet cards are not interchangeable with Ethernet cards. During an upgrade, you will have to incur the cost of new NICs for each node on the network. Two main categories of cards are used in ARCnet networks: Bus NIC (high-impedance driver) and Star NIC (low-impedance driver). As their names imply, they are different in that the Bus NIC should be used on a bus topology and the Star NIC should be used on a star topology. Some newer cards can provide both options.

Tip

When you are choosing cards or devices with embedded adapters, it is important to pay attention to the cabling and connector supported by the adapter or device.

Connecting ARCnet LANs to Ethernet LANs

Because of the different signaling methods used and the different network access techniques, it should be obvious that you can't mix Ethernet and ARCnet nodes on the same cable. However, one of the advantages of ARCnet is that there are vendors who manufacture hubs and other devices that allow you to connect dissimilar cabled networks. For example, there are active hubs that allow for coaxial cable, twisted pair, and fiber-optic cabling. Conversion boxes are also available to bridge between an Ethernet network and an ARCnet LAN. Upper-level protocols, such as TCP/IP, can then be used to pass traffic between the two networks.