Full text of "ARCNET Thomas Conrad-TC6245ARC Card AT Users Guidepdf"

See other formats

Article reprinted by permission of NetWare Connection. 
p or information regarding a free subscription to this 

.agazine, please contact NetWare Connection at the fol- 
lowing address: 

x 19007 

. UT 84605-9007 



ARCnet 



The Rodney Dangerfield of Network Computing 



By Mickey Applebaum 




For the record, let me just admit that I am an 
ARCnet fanatic. I have yet to find a network 
that could not run, and run well, on ARCnet. 
ARCnet is inexpensive, easy to manage and 
install, and virtually as fast as Ethernet in most 
situations. Still, it is losing ground to other 
network protocols, partly because of popular 
misconceptions. 

ARCnet is the only truly multivendor 
standard network interface board 
manufactured today: No matter 
whose ARCnet board you buy, there 
is a 95 percent chance that it will 
work with your network driver. This 
compatibility is a result of 
ARCnet's unique history. (I should 
mention that the Token Ring 
Tropic chipset is beginning to 
standardize Token Ring.) 

In the ARCnet Beginning . . . 

In the late 1970s, Datapoint Corporation 
developed ARCnet, the first network interface 
to transport data from one computer to another 
over flexible coaxial cable (RG-62). Ethernet had 
been developed earlier, but it was used primarily to 
connect large, expensive printers to mainframe 
computer systems. 

ARCnet was also the first network interface to be 
brought to the personal computer in the early 1980s. 
Datapoint developed a specialized communications 
chipset, later changing it to a sealed Large Scale 
Integration (LSI) component, which reduced the cost 
of manufacturing ARCnet network interface boards. 
Datapoint still owns the rights to the LSI component 
and licenses the technology to other ARCnet 
manufacturers, which is why there is a standard driver 
interface. 

When Novell developed NetWare, ARCnet was one 
of the first protocols supported. Novell even had its own 





version of the ARCnet board, the 
RX-Net board — released before 
other Novell network interface 
boards. ARCnet was also the first 
network board to support multiple 
cable specifications and 
topologies. 

The unfortunate part of 
ARCnet's history is that until 
1992, it was never a 
"standard" protocol. In 
1992, the ANSI 
specification 
committee finally 
wrote an ARCnet 
protocol standard. 




How Does ARCnet 
Work? 

ARCnet is a token- 
passing protocol running at 
raw data throughput of 2.5MB 
per second. Token passing means 
that each network interface board is 
allowed to use the cable for a specific 
amount of time. Stations arbitrate their 
insertion points based on each board's 
physical node ID. 

ARCnet boards use physical node IDs 
configured by the network supervisor. Ethernet and 
Token Ring, on the other hand, can use locally 
administered logical node IDs, but their physical node 
IDs are burned into ROM. 

ARCnet boards can be configured for any of 255 
physical node IDs (1 through 255; is used for internal 
timing and addressing functions). When ARCnet 
boards power on or reset, they perform a network 
reconfiguration (RECON), which can be compared to 
an army troop doing a soundoff. Each station sends a 
node ID broadcast and compares it to all the other node 



28 



Guest Column 



NetWare Connection September/October 1993 



IDs. The station with the lowest ID 
arbitrates the soundoff and finds the next 
highest number. The system continues 
from there. 

Token passing also means that 
ARCnet is tolerant of high-volume 
traffic. Since each node takes a turn on 
the cable, the cable does not become 
saturated when it has to transmit large 
amounts of data or data from a large 
number of workstations. Each node 
knows when it can get access to the 
cable and how much data it can place 
on the cable in each timing cycle. 

ARCnet 's performance will be the 
same whether the cable is at 5 percent 



utilization or 85 percent utilization. 
This means that actual node throughput 
on the network doesn't degrade under 
load as it does with collision detection 
network protocols such as Ethernet. 

So What's Wrong with ARCnet? 

With all these things in its favor, why 
isn't ARCnet the most popular network 
protocol? As a true ARCnet devotee, I 
have asked myself that question many 
times. Here are the reasons why I think 
ARCnet is losing its popularity and is not 
being considered for new installations. 

• Reduced cost for Ethernet 
network interface boards 



• No support for Simple Network 
Management Protocol (SNMP) 
systems or management systems used 
by lOBase-T networks 

• A misguided belief that Ethernet 
is significantly faster 

• Limited number of nodes allowed 

The first two reasons can't be 
argued, and for large internetworks that 
require remote management, Ethernet 
and Token Ring provide much better 
options. To give ARCnet its due, 
however, the largest manufacturers of 
ARCnet products (such as SMC and 
Thomas Conrad Corp.) have their own 
ARCnet hub management protocols. 
The biggest problem with these 



ARCnet Cabling and 
Topology 

ARCnet is capable of running on coaxial 
cable (RC-62, 93 Ohm), single twisted 
pair (1 00 Ohm), and fiber optic. ARCnet 
supports twisted pair through the use of 
on-board twisted pair transceivers or 
through the use of coaxial to twisted pair 
baluns. 

ARCnet can run in its native distributed 
star bus, linear bus, and modified star 
bus (multiple linear busses starred off a 
single hub). In its native star bus, 
ARCnet works with a combination of 
active and passive hubs and active 
links. The network interface board in 
each node is considered an active 
device, as are active hubs and active 
links. 

The difference between an active hub 
and an active link is simply the number 
of ports on the device. An active link 
has two ports, and an active hub has 
multiple ports (usually 4, 8, or 16). Any 
two ARCnet nodes can be directly 
connected, connected through a 
combination of active hubs or active 
links with up to 1 active devices 
between nodes, or connected through 
a passive hub or combination of 
passive and active hubs. 



Coaxial 

ARCnet can support the following distances 
for coaxial connections: 

• Star topology, active device to 
active device: 2,000 feet. This can 
be node to node, node to active 
hub, active hub to active hub. A 
maximum of nine active hubs are 
allowed between any two nodes. 

• Star topology, active device to 
passive device: 1 00 feet. This can 
be node to passive hub or passive 
hub to active hub. You cannot 
connect a passive hub to another 
passive hub. You must terminate 
all unused passive hub ports. 

• Bus topology, terminator to 
terminator: 1,000 feet. This must be 
Hi-Z or high impedance boards 
only. You may have up to eight 
nodes connected to each linear 
bus. Each end of the bus must be 
terminated with a 93-ohm termin- 
ation resistor or with a workstation 
using an ARCnet board. This raises 
the limit to 1 nodes. An active hub 
port can be used as a terminator 

on the linear bus. Passive hubs are 
not allowed on a linear bus, 



Twisted Pair 

ARCnet on twisted pair is as easy, since 
there has never been a true standard for 
ARCnet twisted pair. The most common 
distance limitation for twisted pair 
ARCnet is 400 feet with a linear bus 
topology and input and output connec- 
tors on each board. Thomas Conrad 
Corp. has an alternative transceiver design 
that allows for cable distances of up to 
800 feet. Twisted pair ARCnet requires 
that each end of the twisted pair bus be 
terminated with a 100-Ohm termination 
block, which can be a twisted pair active 
hub. With twisted pair, you can have up 
to 10 nodes on each linear bus. 

You can also connect a standard star 
bus coaxial board to twisted pair cabling 
through the use of a coaxial to twisted 
pair balun. In this configuration, though, 
you can only have one station per cable, 
and it must connect to a matched balun 
on a coax active hub at the other end. 

Fiber Optic 

ARCnet on fiber optic can run several 
kilometers between active devices, This is 
the most common use for an active link 
or a two-port active hub, The active link is 
simpler than an active hub and allows the 
maximum distance between nodes. 



29 



Guest Column 



NetWare Connection September/October 1993 



management protocols is that they are 
not interchangeable and, therefore, do 
not allow for a standard management 
system interface such as SNMP or 
NetView. 

As for the third concern, although 
Ethernet has a faster raw data throughput 
(bandwidth) of lOMbit/s, the actual data 
throughput of the current 16-bit Ethernet 
boards is virtually the same as today's 16- 



Token passing 
means that ARCnet 
is tolerant of high' 
volume traffic. 
Performance will be 
the same whether the 
cable is at 5 percent 
utilization or 85 
percent utilization. 



bit ARCnet boards. For example, in my 
own performance tests using PERFORM2 
on my personal network (386SX/16 
workstation, 386DX/25 file server, 
300MB Maxtor 4380S hard disk attached 
to a ProComp F-DCB using the 
DRA.DSK driver with 8MB of RAM 
installed), I obtained the following 
results: 

• Thomas Conrad 6142 8-bit 
ARCnet board in server and 
station: lOOkbit/s throughput 

• SMC PC600FS ARCnet board in 
the server and station: 145kbit/s 

• 3COM 3C503 EtherLink II in 
server and station: 135kbit/s 



• 3COM 3C503-16 EtherLink 11/16 
in the server and station: 
185kbit/s 

As you can see, although Ethernet is 
faster, it is not significantly so. In practical 
terms, the speed difference between 
ARCnet and Ethernet comes down to 
this: saving a 35KB WordPerfect 
document takes .241 seconds with 
ARCnet and .189 seconds with Ethernet. 
Is a savings of .052 seconds worth the 
expense? Furthermore, for my tests I used 
a single workstation and a single server. 
Because Ethernet is a collision detection 
network protocol, it will show signs of 
performance degradation under higher 
packet loads. 

The fourth argument against 
ARCnet — that it allows a limited 
number of nodes — is not really valid. 
ARCnet allows up to 255 nodes on a 
single segment. If you need more nodes, 
you can either add more network 
interface boards to the file server or use 
NetWare routers (ROUTEGEN or 
NetWare Multiprotocol Router) to 
increase the number of segments. 

In a NetWare 2.2 file server with its 
limit of four network interface boards, 
you can have 1,016 workstations without 
using an external NetWare router. In a 
NetWare 3.11 or 4-0 file server with its 
limit of 16 network interface boards, you 
can have 4,064 workstations without 
using an external router. 

Compare these limits to those of 
NetWare itself on a single server — 100 
users for NetWare 2.2, 250 users for 
NetWare 3.11, and 1,000 users for 
NetWare 4-0 — and you can see that 
ARCnet allows many more physical 
connections than you have user slots 
in the operating system. Furthermore, 
you can have up to 254 routers on any 
one segment, and each of those routers 
can have between 762 (NetWare 2.x 
ROUTEGEN) and 3,810 (NetWare 
Multiprotocol Router) workstations 
attached to it. 

ARCnet's Future 

There are two exciting develop- 
ments in ARCnet's future. Datapoint, 
the originator of ARCnet, has 



produced a 20Mbit/s version of the 
ARCnet protocol called ARCnet Plus. 
ARCnet Plus uses the same coaxial 
cabling as ARCnet, and you can use 
both the new 20Mbit/s boards and the 
old 2.5Mbit/s boards on the same 
cabling segment. In addition, the new 
ARCnet Plus hubs will automatically 
arbitrate the speed between the nodes. 
ARCnet Plus can be an easy way to 
obtain a faster network; the only 
network-wide requirement is that you 
have to replace all active and passive 
ARCnet hubs with ARCnet Plus 
active hubs. 

The other exciting development is 
the Thomas Conrad Network Solution 
(TCNS), a lOOMbit/s protocol that uses 
ARCnet's token-passing scheme with 
standard ARCnet and Enhanced 
ARCnet network drivers. TCNS 
supports three different cabling options: 
coaxial, IBM Type 1 shielded twisted 
pair, or fiber optic. With coaxial 
cabling, you can maintain compatibility 
with the installed base of ARCnet 
StarBus systems. 

Using the 16-bit TCNS boards in 
my personal network, I get a 
throughput of 450kbit/s as compared 
to ARCnet's 145kbit/s. Even using the 
Novell EMSNETX and a slow 16-bit 
memory expansion board with QEMM 
as the memory manager, I get 
325kbit/s throughput. Using the same 
16-bit TCNS boards in a 486DX/33 
file server and workstation at my 
office, we consistently get 1300kbit/s 
throughput. 

So spread the word. Low cost, ease 
of installation and maintenance, and 
simple management at the workstation 
and hub level all give ARCnet a place 
in the world of networking for both 
small and large installations. Help 
bring some respect back to this 
Rodney Dangerfield of networking; it 
is still a viable network protocol 
solution, even though lots of folks just 
haven't heard of it. 

Mickey Applebaum works for Uinta 
Business Systems in Salt Lake City, 
Utah. He also spends thousands of hours 
each year on CompuServe as a NetWire 
system operator (sysop) . m 



DATAPOINT Unite ' i States: Datapoint Corporation • 8400 Datapoint Drive • San Antonio, Texas 78229-8500 

Telephone: (210) 593-7000 or (800) 334-9968 • Facsimile: (210) 593-7920 • Document No. 16-2865-001. 10/93.