This chapter is something of an odd duck. It presents and overview of Wide Area Network (WAN) technologies, and provides coverage for the two most common network protocols (TCP/IP and IPX/SPX). Chapter six follows the flow of the class, but is a misfit from the point of view of organizing a textbook. In the next revision chapter six will be divided into two chapters -- WAN Technologies, and Network Protocols. Whether the ordering will change will have a lot to do with whether my presentation of WAN technologies presumes too much a priori knowledge of routers to be placed in with LAN technologies.
That being said, chapter six covers the major WAN technologies in use today. The second and third sections of the chapter delve into Novell's IPX/SPX protocol, and the Internet's TCP/IP protocol. Special attention is paid to IP addressing. Appendix C is provided to define WAN and telecom terminology used in Section I.
A "carrier network" is a large data network owned by a regulated public utility company such as AT&T, Sprint, or MCI. (Sometimes referred to as "telcos".) Some argue that the telcos' monopoly on long-distance data service has acted to retard development and deployment of higher-speed WAN technologies.
[ I ] WAN TECHNOLOGIES
WAN technologies are the technologies used to connect networks together over long distances. Long distances can mean anything from across town to across the planet. Because of the challenges of transmitting high-speed signals over long distances and the capital cost of large carrier networks, the pace of development in the WAN arena has been substantially slower than in the LAN environment. The WAN circuits which were considered high-speed connections twenty years ago are still high speed circuits, while the speed of LAN technologies has increased 100-fold in the same time period.
WAN protocols are hardware protocols, and they fit into the OSI Model in the same way that Ethernet, Token-Ring or FDDI do. Some, like point-to-point leased lines, provide a complete Layer 1 and Layer 2 specification. Others, like Frame Relay provide a Physical and MAC sublayer specification with the expectation of hooking into the 802.2 LLC sublayer. Read Appendix C for definitions of terms and acronyms which I will use in this section.
Current WAN technologies can be divided roughly into three categories: switched circuits, leased lines, and Frame Relay. Switched circuits include dialup modem connections, ISDN, and Switched 56. Leased lines run the gamut from 56K, T1, and fractional T1, to SONET, ADSL, and cable modems. We will cover switched circuits first.
A hot area for dialup connections is remote networking, either in the form of telecommuting (connecting your PC to the office network over the phone line), or connecting to an ISP to access the Internet.
Switched Circuits
Switched circuits are connections that use a telephone switch in the CO to setup and tear down calls on an on-demand basis. Every time you pick up the phone to make a call you are using a switched circuit to "reach out and touch someone". Switched connections are used for a number of tasks ranging from simple dialup terminal sessions, to remote network connectivity, to video conferencing. Some switched services are billed on a per-minute rate for all calls: all of the switched services are billed per-minute for calls outside the local calling area. Charges for switched circuit calls increase dramatically with usage and distance. At some point it becomes more cost effective to use a leased-line or Frame Relay circuit.
The most common switched circuit used for network connectivity is an analog telephone line connected to a PC modem. Present day dialup modems range in speed from 28.8 Kbps to near 56 Kbps. Dialup modem technology has reached a peak in terms of speed. Current high speed modems rely on heavy built-in data compression and good quality connections (no line noise, low signal loss) to achieve speeds above 19.2 Kbps. Because of the sampling rate of analog-to-digital converters in the PSTN, dialup modem speeds have an upper bound in the area of 53 Kbps. This speed limit is further enforced by FCC regulation.
Switched 56 is a true 56 Kbps connection, used primarily in business for network-to-network connectivity or video conferencing. Once the lines and equipment have been installed, Switched 56 calls are made just like a regular telephone call. Area codes in the 700 range are usually Switched 56 numbers.
The other current switched circuit technology is ISDN, or the Integrated Services Digital Network. ISDN lines are made up of at least three channels. An ISDN line you would use in a home or small office is a Basic Rate line, also referred to as ISDN BRI (Basic Rate Interface). An ISDN BRI line has two B channels for carrying data, and a D channel for signalling. The B channels carry 64 Kbps each. A B channel can be used for a data connection, or a telephone line. The D channel operates at 8 Kbps and is used to communicate with the CO switch to setup, tear down, and maintain calls. In practical terms this means that you can be connected to your ISP on one B channel while talking on the phone or receiving a fax over the other B channel. When used for data service, the two B channels can be bonded together to provide a 128 Kbps connection. Whether or not an ISDN subscriber can use channel bonding is determined by both his equipment and the equipment of the party he's calling.
ISDN PRI (Primary Rate Interface) consists of twenty three B channels and one D chanel, all carried over a T1 circuit (covered below). ISDN PRI lines are used by corporations to connect multiple telecommuters or by ISPs to provide high-speed Internet access to multiple subscribers.
Deployment of ISDN, which was developed years ago, has been slow due to the high cost of outfitting a CO for ISDN service. The explosion of public interest in the Internet has helped speed both deployment and acceptance of ISDN, but its future is murky at this point. ISDN had the promise of providing a revolutionary new type of "telephone" service -- pulling together both voice and data service on the same line. But, many LECs are now waivering in mid-stride with ISDN service half-way deployed in their networks. Hesitance on the part of the LECs is due to the emergence of newer, higher speed, technologies such as ADSL and cable modems (both covered below). Heavy users of the Internet already want more speed than ISDN can provide. Attempting to project the future of ISDN at this point in time would be pure speculation on my part. Time will tell whether ISDN will go directly from rising star to has-been without ever having it's moment in the spotlight.
Leased Lines
Any one of several forces might push a business away from switched circuits and onto leased line or Frame Relay service. When long distances are involved, switched service rapidly becomes more expensive than tradition leased line service when useage rises above an hour or two per day. Some applications may not tolerated the delay inherent in dial-on-demand routing (using a switched circuit between two routers for a WAN connection). The fastest switched service -- ISDN BRI -- topps out at 128 Kbps: when more bandwidth is required, something other than switched service is needed.
Traditional leased line service is centered around the technology of the T1. The T1 circuit is a 1.5 Mbps point-to-point line carrying synchronous digital data. T1s use time division multiplexing (TDM) to divvy up the 1.5 Mbps into 24 channels, each carrying either 56 or 64 Kbps. Each channel of a T1 can be used to carry a single voice telephone call, or a data connection. A T1 can be used as one large unit for a single 1.5 Mbps data circuit, or channels can be grouped together to create a data circuit operating at a fraction of the T1's 1.5 Mbps bandwidth.
A T1's channelization can be put to several uses at the same time. A T1 line going into an office might be configured to use twelve channels for incoming and outgoing toll calls (connected to the office PBX), four channels for a 256 Kbps connection to the corporate headquarters, a pair of two channel 128 Kbps connections to other field offices, and four channels left idle for future expansion. The device that handles the separation and combination of T1 channels is called a channel bank. The diagram below illustrates the connections described above.
You can get leased line service at sub-DS0 speeds, either 9.6Kbps or 19.2Kbps. However, because the circuits will be carried from CO to CO on DS0 lines, the LEC has to provide special equipment in its COs to handle the lower speeds. The result? Sub-DS0 service usually costs a few dollars more per month than a simple 56K or 64K line.
T1 technology was developed by AT&T at its Bell Laboratories. Much of our WAN terminology still reflects its AT&T origins. In AT&T's nomenclature a T1 is a "DS1" circuit. A DS0 is a single 56 or 64K channel. At the DS3 level (45 Mbps, the top of the DS scale) we overlap with the bottom of another scale: the "OC" (Optical Carrier) scale. DS3 service equates to OC1. The OC levels march upwards in roughly 51.5 Mbps steps from OC1 at 45 Mbps to OC3 at 155 Mbps and beyond. The table below shows the common classifications of service, the speed of that service, and the technologies which use that service.
The cost for a WAN circuit increses exponentially with speed and distance. For instance, a 56K line from an office in one town to an office in the next might cost $230 per month, while a full T1 might cost $800 per month. A T1 circuit covering any appreciable distance can easily cost over a thousand dollars a month. T3 service, even between two points in one town, can cost over $100,000 per year. The staggering cost of high-bandwidth WAN connections does three things:
- Creates a growing disparity between LAN speeds and WAN speeds. When most companies deployed 10M Ethernet or 4M Token-Ring networks they were using 1.5M T1s. Now, those companies are deploying gigabit Ethernet networks, and they are still using 1.5M T1s.
- Relegates high-speed WAN technologies to use by only the largest corporations, ISPs, and the carriers themselves.
- Drives development of cheaper multi-megabit technologies by groups
such as the cable industry.
As of early 1999, Asyncronous Transfer Mode (ATM) is still a technology in flux. Initiaily, is was going to become the One True Network Technology, and replace everything else from the desktop to the WAN. This hasn't happened. Several things have gotten in the way of ATM's acceptance by the networking community. First, it took a long time for ATM standards to settle down: no one wants to buy a product that probably won't work with a product from a different vendor. Because ATM is based on switched connections at Layer 2 (somewhat like the telephone system), and not shared media (like Ethernet or Token-Ring) its has proven very difficult to get ATM to fit cleanly with upper layer protocols. ATM has been around since the early 1990s, but native TCP/IP over ATM is a development of the last two years. And, the cost per-port for ATM hardware still remains very high.
ATM is a compromise technology. It uses a switch at its core, requiring call setup and teardown. ATM does not use frames like Ethernet or Token-Ring, it uses 53 byte "cells". This diminutive size is a compromise between the voice and video crowd who wanted an even smaller cell (better to handle their type of traffic) and the data crowd who wanted a much larger cell (more akin to the 1500 byte Ethernet frame). ATM can operate at speeds from DS1 to OC12, with common implementations at 25 Mbps, 155 Mpbs, and 622 Mpbs. ATM is an outgrowth of the same committee work that produced ISDN, and seems to have the same dubious taint.
ATM remains a niche technology. The major carriers are the primary consumer of ATM equipment. AT&T, and MCI are building large ATM infrastructures to carry the next generation of voice and data services. But, at this time, their ATM infrastructures remain hidden from customer view: ATM serves as the POP-to-POP carrier of services that actually reach the customer, such as T1s and Frame Relay. ATM has carved out something of a role for itself in the very high-speed LAN arena. But, outside of networks that carry a high proportion of streaming video, ATM is being challenged by the emergence of 802.3z Gigabit Ethernet (GbE). Where ATM will be in five to ten years is anyone's guess.
The Small Office, Home Office (SOHO) market is a rapidly growing arena that has attracted considerable attention from networking vendors and service providers.
The technologies discussed above are primarily used by businesses to connect to their plants and offices, to other businesses, and to corporate ISPs. Two technologies are emerging on the home front which can bring megabit speed, or higher, service into the home or small office. The two are cable modems, and (A)DSL.
A cable modem is essentially a speed matching bridge. One one side you have a PC with an Ethernet card (or a hub and several PCs) and on the other side you have a connection to your regular coaxial cable wire. Cable modems operate in the megabit speed range; all of the subscribers served by the same cable head-end equipment are on the same segment. This is a shared media technology, like traditional Ethernet, so the throughput you recieve is affected by the amount of activity generated by other subscribers on the segment. Cable modems are symmetric -- they have the same data rate receiving from the head-end as they do transmitting to the head-end. Cable modem service is not available in all areas. However, where it is available, it does offer a low-cost means of making a high-speed connection to the Internet. Cable modems connections are not really suited to business applications because the useage agreements generally prohibit subscribers from connecting "servers", be they email, web, or whatever, to the cable modem service.
See what John Perry Barlow had to say about asymmetric services in his May 1995 article in Communications of the ACM. The article is in Adobe PDF format.
Digital Subscriber Line (DSL) service has been available for some time. High-speed Digital Subscriber Lines (HDSL) have been used to carry T1 circuits over short distances (eg: within a town). A new development, the Aysmmetric Digital Subscriber Line (ADSL) could provide telcos with a counter to cable modems. ADSL lines are point-to-point connections like a T1 or 56/64K leased line. The difference between ADSL and traditional leased line technologies lies in the speeds. ADSL supports high-speed transfers from the CO to the subscriber (8 Mbps), but supports much slower transfers from the subscriber back to the CO (in the area of a few hundred Kbps). This makes ADSL a WAN technology aimed at information consumers, like the average web surfer, and not aimed at information producers, like corporate web sites.
With cable modems and ADSL the WAN situation looks good for the home user or the SOHO client. The WAN situation for corporations needing long-distance connections and symmetric bandwidth is not as cheery. While prices for T1 service have come down in recent years, higher bandwidth and lower costs still seem something of a dream for corporate network managers. Research is ongoing, however. The map below shows the Abeline Project's Internet2 network. The Abeline network connects research institutions in 31 cities at OC48 speeds (2.4 Gbps).
Frame Relay
Frame Relay is a WAN technology like the others in the previous section, but because of its prevalence and technical intricacies, it warrants discussion in its own section. Frame Relay's popularity is driven by its utility as a cost-saving tool for WAN managers.
The really expensive part of a point-to-point WAN circuit is the bandwidth in the carrier's network that is dedicated to carrying your traffic. With a T1, the carrier has to dedicate facilities from end-to-end to carry 1.5 Mbps for you, whether you are using the circuit at the moment or not. Frame Relay allows you to drive WAN costs down by attacking dedicated bandwidth in two ways.
Frame Relay uses traditional leased lines to carry traffic from your premise to the carrier's POP. But that's it. If you have a cross-country Frame Relay link you'll only pay leased line charges from your office to the carrier's nearest POP, and then from the POP on the far end to your remote office. That's the first way Frame Relay saves money - by reducing the number of miles of leased line used in the connection. Once your traffic has reached the carrier's POP it enters the carrier's Frame Relay Cloud. Once in the cloud, your data shares the carrier's network bandwidth with other subscribers' data.
Two things regulate the speed of a Frame Relay connection. The first is the port speed: the speed of the leased line connection from your location to the POP. This defines the upper bound for how fast the link can operate. The other is the Committed Information Rate (CIR). The CIR is the minimum amount of data that the carrier agrees to carry over the link at any given time. Generally the CIR is half of the port speed, though you might elect to have a lower CIR. The CIR and sharing facilities in the cloud are the other cost saving mechanism in Frame Relay: you reserve only a portion of the bandwidth you would for a traditional leased line connection, and the cost of the supporting facilities (the cloud) is shared among many subscribers.
The port speed and CIR define the maximum and minimum data rates for a Frame Relay link. Having a variable data rate is something new: all of the network technologies we have examined previously have had fixed data rates. A carrier's Frame Relay network generally does not have enough bandwidth to support all of its users if they all tried to transmit at their maximum data rates at the same time. When a subscriber transmits at a rate higher than his CIR, the Frame Relay switch in the POP marks all the frames that come in above the CIR as discard eligible. Transmitting above your CIR is called bursting. If the Frame Relay network becomes congested due to heavy traffic, a switch will drop discard eligible frames in favor of forwarding non-discard eligible frames. In essence, Frame Relay's traffic control mechnism works to keep the committed traffic flowing by dumping burst traffic when it has to. It's up to the transmitting and receiving stations on either end of the link to recognize that frames have been discarded and to resend their data.
To the end stations (usually routers) on a Frame Relay connection, the link looks like a straight-though connection. This is just a logical appearance however. Frame Relay, as it is implemented in the US, uses Permanent Virtual Circuits (PVCs) to give the impression of a point-to-point connection between the POP where a frame enters the cloud and the POP where it exits. Behind the scenes in the cloud, this isn't the case. The diagram below shows a Frame Relay link. The PVC is the broken orange line. The Frame Relay network can route frames through the network on any path from POP to POP to avoid congested links. The other colored lines within the cloud show the actual paths that frames from the two routers might take as the network forwards them through the Frame Relay cloud.
Because a PVC is a logical connection, more than one PVC can be passed over a physical connection between the Frame Relay switch in the POP and the subscriber's router. This permits construction of "hub and spoke" networks where the hub has only one (large) leased line connection into the cloud and the smaller sites (the spokes) have smaller connections to the cloud. This results in a further reduction in the number of leased line connections and (expensive) router ports, compared to a traditional large leased line WAN. A hub and spoke network might be constructed to link up the Point of Sale (POS) terminals in a number of department stores with the inventory control system in the corporate headquarters.
Frame Relay is a vital, and widely accepted WAN technology which should have a long life ahead of it. Short-range solutions such as ADSL are not a threat to nation-wide and global Frame Relay carriers, and ATM is being used much more frequently to carry Frame Relay traffic than to replace it.
[ II ] IPX/SPX
Internetwork Packet eXchange (IPX) and Sequenced Packet eXchange (SPX) are the creations of Novell Inc. Novell created IPX/SPX as part of its NetWare operating system. IPX operates at the Network and Transport layers. Its job is the reliable transport across the network. To do this, IPX handles network addressing and routing. SPX operates at the Transport layer and handles packet sequencing in streams of data. SPX is a connection-oriented protocol. Upper layer services on a NetWare network use IPX/SPX for carrying file & print, login, and remote console traffic. IPX/SPX's fit into the OSI Model, and other upper and lower layer protocols is shown below. Most of this section will be devoted to discussing IPX.
IPX Addressing
Novell designed IPX to be simple, robust, and low-maintenance. The greatest simplification built into IPX comes in the form of station addressing. The network portion of a station's Layer 3 address is set by the administrator when he or she sets up a NetWare server or IPX router on the network. Stations learn what the network portion of their Layer 3 address is at boot time when they first attach to the network. For the host portion of the Layer 3 address, IPX uses the station's Layer 2 MAC address. By using this method, network addresses need only be set in a few places (servers and routers), and workstations automatically handle setting their Layer 3 address -- without administrator intervention.
The IPX address space is quite large. How roomy will become apparent when you read about IP addressing in the next section. The network portion of an IPX address is 32 bits long. The host portion of an IPX address is taken from the station's 48 bit MAC address. The total address is 80 bits long. Theoretically this would permit you to build a network of 4.3 billion segments, with a total of 281 trillion stations. Obviously, this is unattainably large. But, in practical terms it means that network administrators have no fear of running out of IPX network addresses on even the largest networks.
IPX RIP and SAP
Automatic addressing does however, generate a downside for IPX, particularly on large networks with hundreds of servers. Because addresses can change quite easily, Novell needed to build a protocol to identify where servers are on the network, and what services are available from them. The result is SAP, the Service Advertisement Protocol. SAP has been quite rightly branded a "chatty" protocol. Servers on an IPX network send out regular SAP updates every two minutes to insure that stations on the network know where they are. On a very large network this can create a river of SAP traffic that never stops flowing. In a large WAN environment precious WAN bandwidth can be wasted carrying SAP traffic. The existance of SAP filtering software on most routers is evidence of the seriousness of this problem on large networks.
Perhaps because of its development relatively early on in the era of networking, or because of the desire for a simple (and thus easy on the processor) protocol, IPX was not built with packet fragmentation and reassembly capability. With Novell's original distance vector routing protocol (RIP or Routing Information Protocol), IPX defaulted to using the smallest possible frame size (ARCNET's 512 bytes) to avoid any situation where a packet might need to be fragmented. This was not the most efficient way to utilize a network technology like Ethernet or Token-Ring. With Novell's second routing protocol (NLSP or Novell Link State Protocol), IPX still can't perform packet fragmentation and reassembly. However, NLSP can determine what the smallest frame size actually being used on the network is, and IPX will use that. So, on a mixed Ethernet and FDDI network using NLSP as its routing protocol, frames will at least be 1500 bytes long. Because NLSP is a link state protocol it is much less chatty than RIP. These two improvements have removed a serious criticism of early IPX.
[ III ] TCP/IP
TCP/IP, the Transmission Control Protocol / Internet Protocol, is quite probably the single most used network protocol, period. TCP/IP is the language of the Internet, and as such is spoken by millions of systems seven days a week, twenty-four hours a day.
What is more properly referred to as the "TCP/IP Protocol Suite" is the product of research initiated in the 1970s by DARPA, the Defense Advanced Research Projects Agency. DARPA's goal was to build a data network with enough robustness and survivability to carry on functioning in the aftermath of a nuclear war. From that very ominous beginning has come the Internet. Initially a network linking colleges, universities, and a select few companies, the Internet has grown to reach almost every nation on Earth. The engine that drives it all is the layers of protocols in the TCP/IP suite.
Like Novell's IPX, IP handles routing of packets across the network. Unlike IPX, IP has packet fragmentation and reassembly built in, so it will always use the largest (most efficient) frame size available to it.
TCP is another connection oriented protocol. Like Novell's SPX, TCP works by establishing a connection with a remote machine, managing an orderly exchange of packets, and then shutting down the connection. Many of the upper layer services like email, web browsing, and file transfer ride on top of TCP.
A newer Transport layer protocol has been gaining popularity in recent years: UDP, the User Datagram Protocol. UDP is a connectionless protocol. It is used for broadcasting (and multicasting) data across a network where there are multiple recipients and proof of reception isn't needed. MBone, a protocol used for broadcasting music over the Internet uses UDP. Other protocols for streaming audio and video content to multiple users also ride on top of UDP. For these applications, using UDP (which does not guarantee delivery) generates less overhead for the sending station to deal with.
IP Addressing and Classes
Compared to IPX's 80 bit address space, IP's Network address space is quite small. In fact, the diminutive size of the current version of IP (IP v4), may result in the exhausting of available network addresses on the Internet. IP uses a 32 bit address, divided up into four eight-bit bytes, or octets. The front portion of the IP address indicates the network portion of the address, and the rear portion indicates the host portion of the address. Where the dividing line is, between network and host parts of an IP address, is not fixed. Unlike an IPX address which has a 32 bit network portion and a 48 bit host portion, the number of bits in an IP address devoted to each part is flexible.
The primary mechanism for determining where the dividing line is between network and host parts of an address is the address' class. There are three classes of IP address; A, B, and C. In a class A address, the first octet identifies the network number, and the trailing three octets identify the host. In a class B address the octets are split 50/50: the first two octets identify the network segment, and the trailing two identify the host. Predictably, a class C address has the first three octets of the address for the network portion, and the last octet for the host portion.
To render IP addresses in a form humans can read and understand we write the value of each octet in decimal form (base ten), separated by a period. This is known as dotted decimal notation. In decimal notation, the possible values for an eight bit byte range from 0 to 255. An IP address in dotted decimal notation looks like: 207.22.201.199 Which class an address belongs to is determined by the value of the first octet. The table below shows the values of the first octet for each class.
Older implementations of IP could not properly handle a network or host address whose bits were all zeros or all ones, so these addresses are not used.
In a class A address, the first bit is always zero. This means that the remaining seven bits of the byte are used to indicate the network number. At first glace this would give 128 possible values (0 - 127), but decimal 0 (meaning all bits of the byte are set to zero) is not used in practice. 127 is a loopback address, so the useable number of class A network addresses is 126. On the Internet, all class A addresses were assigned long ago. The bits for the maximum and minimum values are: 00000001 and 01111111 respectively.
In a class B address, the first two bits are always one and zero, respectively. With all other bits set to zero, this means the minimum value of the octet is 128. With the right-hand six bits all set to one, the maximum value of the octet is 191. The bits for the min and max values are: 10000000 and 10111111. As of this time all of the Internet's class B addresses have been assigned to colleges, universities, governments, and businesses.
The leading bit pattern for a class C address is 1 1 0. This leaves the right-hand five bits to set a value for the octet between the minimum of 192 and the maximum of 233. Why is the maximum 223? Add up the values of the five right-hand bits, and the first two leading bits. The result? 223. Another way to reach this answer is to take the maximum value for an eight bit byte (255) and subtract the value of the bit that must always be zero in a class C address (64). The result is, of course, 223. This maps out to max and min bit patterns of: 11000000 and 11011111. Thankfully we're not yet at the point of exhausting the Internet's supply of class C addresses.
Visit the IANA and the InterNIC (run by Network Solutions) to learn more about Internet addressing and naming.
When IP was first developed, it proved cumbersome from an administrative standpoint in large networks. Each system had to be manually addressed, and strict records had to be kept about who had which address. Also, to access a system you had to remember its IP address. While some of us live and breathe these things and can rattle off long strings of IP addresses from memory, this isn't how most of the computing public would like to work. Two systems have made life much easier for the network administrator and the end user.
The first is DHCP -- Dynamic Host Configuration Protocol. DHCP permits a station to start up on the network and request an IP address from a central server. In addition to providing an address, DCHP can also provide the IP address of the local router and DNS servers (see below). This makes adds, moves, and changes, on an IP network much easier. The user can simply plug in at her new location, boot up her PC, and away she goes. The network administrator doesn't need to get involved.
The second is DNS -- Domain Name Service. DNS Servers maintain a distributed database used to match systems names to their IP addresses. This removes the burden of remembering IP addresses from the user, and leaves them with the simpler task of remembering a name. Using DNS, users can reference a database server with the IP address of 220.176.50.2 as "dbserver", or "fred", or whatever name is setup in the DNS database. Each time you type a name like "www.microsoft.com" into your web browser, your browser is using a DNS server to match the name to an IP address.
IP Subnets
Until IP v6 reaches general availability, we must still contend with IP v4's 32 bit address space. The problem is with the number of network numbers available for use, not the number of host addresses. A technique called subnet masking can be used to extend the number of networks available.
Each class of address has a mask normally associated with it. The chart below shows the "standard" masks for each class. The value of each octet of the mask indicates how many bits of the IP address to use for the network portion of the address. A byte with all eight bits set to one equals 255, so when the value of an octet in the mask is 255 it means "use all the bits of this octet for the network address." If you match up the standard masks for each class to the chart showing the dividing point for each class (above) you will find an exact match. Where ever the mask has an octet wth a 255 value, that octet of the IP address is used to indicate the network number.
You can subdivide the host portion of an IP address into "subnets" by setting a mask which specifies more bits for the network portion of the address than is normal for the address' class. For example, a class B address has a standard mask of 255.255.0.0 This tells us that all of the bits in the first two octets are used to indicate the network number. By specifying a value greater than zero for the third octet of the mask, we can start using bits from the third octet of the IP address to indicate the network number.
132.41.x.x is a class B network number. Normally it has a mask of 255.255.0.0 Let us set a mask of 255.255.192.0 By the rules of reading a mask, the new mask tells us to use all the bits of the first two octets for the network address and the first two bits of the third octet. (192 = 128 + 64, 128 and 64 are the values of the first two bits in an octet.) This means that from one network number (132.41) we have now created four sub-networks, or subnets. The chart below shows the bit patterns for the third octet for each of the four subnets, and gives the resulting IP addresses. The portion of each octet used for expanding the network address is shown in red. The part left for specifying the host address is shown in black.
| 00000000 to 00111111 | 132.41.0.x to 132.41.63.x |
| 01000000 to 01111111 | 132.41.64.x to 132.41.127.x |
| 10000000 to 10111111 | 132.41.128.x to 132.41.191.x |
| 11000000 to 11111111 | 132.41.192.x to 132.41.255.x |
For a host with the full IP address 132.41.184.207 the bit pattern for all four octets is (network portion in red, host portion in black):
Using this technique you can greatly expand the number of network addresses available to you without having to request a new address from the IANA. This is a win-win, in that you can get the addresses you need for new network segments and the Internet community at large does not have to give up a whole new class C address (class A and B addresses are not available). For example, an ISP providing connections to customer sites via Frame Relay only needs two IP addresses for each Frame Relay link (one for the router on each end). This would be a waste of a class C address. By using a mask of 255.255.252 the ISP can stretch one class C address to cover 64 links. Similarly, a university with a class B address might use a mask of 255.255.255.0 (the class C mask) to divide their one network address into 256 subnets, each with up to 255 stations.
[ IV ] CONCLUSION
Chapter six covered two divergent areas of networking: WAN technologies and upper layer network protocols. We have surveyed the current WAN technologies, and explored T1s and Frame Relay in detail. In the network protocol sections we have discussed IPX/SPX and TCP/IP, with particular attention paid to network addressing in both. Chapter six finishes with an in-depth explanation of IP subnets.
[ V ] SELF CHECK
- You are the network manager for a company which owns a dozen gas
stations in the county. The stations need to upload transaction information
once a day to a central database at the franchise headquarters. The average
daily upload is less than 1MB. What WAN technology would you propose to
provide the lowest-cost connectivity?
- Same scenario, but now the franchise has installed pay-at-the-pump
credit card systems which require realtime validation of customer credit
cards. How might this change your proposed network?
- How many data connections can you support over a single T1? Over
an ISDN BRI line? (The ISDN question is something of a trick!)
- A corporation has offices in Detroit, Chicago, Dubuque, and Kansas City.
What is the minimum number of WAN connections needed to give each site
a redundant connection to the corporate network?
- Describe the role that ATM plays in wide area networking today.
- What was the complaint against IPX/RIP, and how is it addresed by
IPX and NLSP?
- What class are each of the following the IP address in? 193.77.35.253,
14.246.245.244, 137.1.1.5
- You have a class C network address and would like to subdivide it
into eight subnets. What mask would you use?
- You have a class A network, and you would like to subdivide it. What
address mask would you use to specify three octets of network address
and one octet of host address?
- Which network protocol, IP or IPX supports packet fragmentation
and reassembly?
|
[ CH 5 ]
[ TOC ]
[ CH 7 ] |
|
1999,2000 Shipman | Created 2-21-99 | Updated 2-6-00