Internet Connections

Bit rate is the number of bits per second that is transferred. Baud rate is the number of times per second the signal changes between states. Baud rate and bits per second are not necessarily the same because each baud could be more or less than one bit. Most high speed connections use a "multiple-state modulation" to transfer more than one bit per baud. This is where instead of each baud representing a 1 or a 0 depending on it's current state, it could be divided into frequency ranges, each range is equal to a sequence of bits.

Telephone lines are only analog and were only designed with analog signals in mind. This is the limiting factor for modem connections. All dial-up modems use the same channels as standard telephones do. The way in which telephone offices process signals limits the connection speed to 64kbps. Because of signal distortion and interference, the maximum speed in which modems can connect is 53kbps. For digital computer signals to use analog phone lines, a modem has to be used. Modem's got their name from what they do. They are a modulator / demodulator. Modems convert the computer's binary digital signal into electrical waves. The computer's 1's and 0's are converted to electrical peaks and valleys.

Modems use the phone line to send an analog signal to the internet service provider. This is much like a regular phone call, it still consumes a phone line, and the call still has to be dialed. In this case, instead of using a telephone to dial, the modem dials, and communicates in the same frequencies that a human voice would. When the analog signal reaches the ISP, the signal is converted back to digital form. The ISP in most cases has a digital phone line to the public switched telephone network (PSTN). If the ISP does not have a digital line to the PSTN, then the signal is again converted to analog. The signal is sent through either the ISP's digital or analog signal to the PSTN, which has a direct connection to the Internet backbone. The Internet backbone is a group of high bandwidth lines, either fiber optics, multiple digital lines, or satellites, which are used to get signal to its destination PSTN.

Standard DialUp Modem

Dial-up connections are the slowest, but most popular connection to the Internet. The first dial up modems were limited to 9600 baud, but went through advances to 28.8 k baud, and then 33.3k baud. And 33.3k baud is were it settled. But then new advancements came in 1997 when 2 competing 56k baud transmission specifications were first introduced. There was Rockwell's 56kflex and US Robotics' 56kX2. These both were capable of 56k baud download and 33.3k baud uploads, and only worked when the ISP had a digital line to the PSTN. Both of these had their advantages, and became equally popular. Because both of these standards were incompatible, modems were limited to either one or the other, depending on the brand of modem that was purchased. But lucky for the end user, most ISP's decided to upgrade their technology to support both standards. Later, to establish some conformity, the 2 56k technologies were merged into what is known as 56k V.90; which took the positive aspects of each previous version, along with a new compression algorithm, resulting in a faster and more reliable connection.

The successor to the V.90 standard, the V.92 standard was recently agreed upon, and modems are expected on the market in the fall of 2000. The V.92 transmission specification has a faster handshaking, 25% more efficient compression, and a 48k baud upload speed.

Dial-up modems have their advantages. They are able to use standard phone lines, use inexpensive hardware, and can not only connect to the Internet, but can double as phone appliances, such as a fax machine, answering machine, or telephone. The disadvantages of a dialup connection are it has slow transfer rates, high pings, and they consume a telephone line when in use.

ISDN is still a dial-up connection, but it has a few key differences. ISDN is a digital connection that is used over regular lines. An ISDN line is made of two regular copper telephone lines, each with their own phone number so they can be used with regular telephones. Because ISDN is dial-up, it means that the telephone line can not be used by both a ISDN connection and a telephone. The advantages of being digital is that the connection is 64k for each line, connect rates are always at 64kbps, ISDN signals can be processed by routers, bridges and hubs. The disadvantages are that special offload circuits have to be installed into the telephone line, and like DSL, ISDN is range limited. ISDN is limited to 18000 cable feet for a connection to be possible. ISDN connections can use either 1 or 2 phone lines using inverse multiplexing. While using only one line, the maximum speed is 64k, and for 2 lines it is double at 128k. A typical 128k connection uses two 64k bearer (B) channels and one 16k delta (D) channel for a controller. Because ISDN needs a controller, in some areas that have older telephone switching equipment, the controller has to share the same bandwidth as the data. This results in a 56k single line connection and a 112k double connection. ISDN never caught on for consumer level because it suffers the same range restrictions as DSL, but offers only a fraction of the speed. Although this connection is only 64k or 128k, because it is a digital signal, throughput is always 8k or 16k respectively for both uploads and downloads. The digital signal also allows for lower ping times, around double that of a cable or DSL connection.

DSL was developed to be broadband connection that could compete with cable. DSL uses standard phone lines to send analog signal link a regular modem, but uses the unused frequencies for greater bandwidth. Regular phones use the frequencies between 0-4KHz, while DSL often use the frequencies between 26KHz and 1MHz. DSL signals are filtered out before the PSTN processing occurs. This means that DSL modems are capable of using all of the frequency ranges like a cable modem is, except for the small portion that a standard phone uses. Because DSL modems do not use the standard telephone or standard modem frequencies, the phone line can be used simultaneously between the two. And because DSL signals do not have to travel through PSTN, they do not have to dial in to make connections.

DSL is dedicated bandwidth, unlike cable, and offers multiple speed grades, depending on the users distance from the PSTN telephone office. High bandwidth in excess of 2Mbps are capable with 12000 miles of the PSTN, but available bandwidth diminishes the farther away from the PSTN the user is. The maximum distance that DSL is able to operate at is 18000 cable feet from the PSTN. This is because, unlike coaxial cable lines, phone lines were not designed to handle high bandwidth signals. The greater distance the signal has to travel, the more interference it will have. Interference disrupts the high frequency ranges, so they may become too distorted for use at long or even moderate distance.

DSL comes in many different variations, the most popular is Asynchronous DSL, which offer different speeds for upload and download, making is more economical for the average user. DSL connections are cable of very low pings, in the ranges of sub 10ms.

Cable connections offer broadband connections through the existing cable television lines. With special equipment, cable providers are able to send and receive analog signals using high bandwidth coaxial cable. The computer's digital signals are converted to analog signals by a cable modem, and broadcast through the cable lines to a Cable Modem Termination System (CMTS), which is connected to the Internet backbone. Unlike telephone line modems, cable modems are not limited to only audible analogue signals, but instead can use the entire frequency range for signals, resulting in a lot of extra bandwidth. The signal from the cable office is approximately 300 to 450MHz depending on the company. This signal is divided into 6MHz increments, each increment is a respective television channel. Cable modems use the unused signals for receiving and transmission of data.

Cable modems, like LAN connections and cable television, broadcast their signals and share bandwidth. Signals are not direct lines, like a telephone line, but a collective group. This means that all cable users share the same bandwidth. To prevent this from being a major problem, cable companies have grouped users to nodes, each node has their own dedicated CMTS and limited users. This prevents large fluctuations in bandwidth, but still doesn't solve the problem. Each user on the node still shares the same bandwidth, and the more users on the node, the less overall throughput each individual has. To prevent an individual user from using all of the bandwidth on a node, cable modems are transfer capped. This is a limit imposed which can vary from company to company. Most cable modems are capped at between 500kbps and 1000kbps download, and between 200kbps and 500kbps upload. Because bandwidth is still shared on a node, during peak hours cable bandwidth is significantly lower than the imposed cap. Cable modems are capable of ping times at least 10 times lower than dial-up modems, and equal or a little higher than DSL.

Unlike cable and DSL, satellite Internet service is available almost everywhere. It operates by having all downstream information sent to you by satellite through your satellite dish. All upstream information has to be sent by a standard dial-up connection, because the satellite receiver is a one way connection only. This puts some limits on the overall usefulness of satellite. It doesn't have the ability to be always connected, and it consumes a phone line when it is. It also has limited bandwidth. While the download speeds can range from 400kbps to 400Mbps, upload speeds are capped at 33k.

Satellite doesn't offer the low latency pings that other broadband connections have, and anything that needs uploading will take close to the same time as for a regular dial-up. The price is comparable to cable prices, and there are also deals if you use the satellite ISP as your television satellite vendor. The dish is slightly larger than the standard mini-dishes, and speeds will slow down whenever weather is stormy.

Inverse multiplex connections, other known as multilink, channel aggregation, channel bonding, shotgun, or load balancing, is a way to combine multiple slow connections into one faster one. Almost any connection can be inverse multiplexed, but only the slower connections really need it. Regular modems and ISDN connections are the usual favorites. Standard ISDN uses a double phone line to achieve 128k connections, and was built with bonding in mind. Standard modem connections later took IDSN's bonding technology and started to use it, because standard modems have a lot more to benefit from bonding. This is because multiple phone lines are cheap, and unlike ISDN, do not require the expensive offloading circuitry is installed. In some areas where cable availability has not yet reached, and ISDN and DSL connections are out of range, bonding modems is the only choice for a high speed connection.

One of the first bonding modems on the market was Transend's 67.2K modem. This bound two 33k connections. Later onto the market there were products that featured 2, or even 3 56k modems on one expansion card. This produced up to 112k and 168k connections respectively. Through software, up to 128 modem devices can be bonded together, but to find a computer that will support that many cards is another task. The most popular software is Diamond Multimedia's Shotgun software. This software will bind 2 modems, even of different speeds into one connection using 2 phone lines. Another feature of this technology is the ability to drop the second line when an incoming phone call is detected. The Windows operating system also supports binding up to two modem connections through software.

Although binding multiple connections can increase bandwidth, it is not without its downsizes. First, your ISP must also support this feature. Second, each modem requires its own phone line. And third, ping times are the same as they would be for a single modem.

Computers can share one Internet connection, whether it is a dial-up account, cable, ISDN, or DSL. Sharing eliminates the need for multiple services, and multiple phone lines for dial-ups. By allowing multiple computers to share a single connection, the bandwidth is divided between all of the computers, in other words; shared, allowing each of them to use it. Sharing the Internet connection doesn't increase bandwidth, all of the bandwidth is distributed between the computers. This means that while multiple computers are using a shared connection, each computer's total available bandwidth will be a fraction of what it would be on an unshared connection.

Internet sharing works well with all Internet and network connections, and is a perfect use for high speed services, like DSL, where one user has a hard time using all of the bandwidth alone.

To have either gateway or proxy sharing, all of the computers have to be first networked together, for example using 10Base-T connections. With an existing network, one computer is put in charge of the Internet, and simply allows others on the LAN to use it.

A gateway is simply a router that will check each packet to see if it is a LAN packet, or a WAN packet.

Proxy connections can be made by either using hardware or software products. Hardware proxies are more oriented toward the business market, and software solutions are more for the average home user.

Hardware
Hardware proxies do not use a computer to operate a proxy, all work is done inside an intelligent hub. Features for hardware proxies include : Network Address Translating for connecting of dissimilar network protocols, DHCP which is used for dynamic IP configurations, DNS for server locations, and firewall securities for packet filtering. Many hardware solutions are not only an intelligent hub, but also have an integrated Internet connection device. This means the hub is not only a proxy server, but also a modem. This type of hardware proxy replaces the extra computer. The Internet connection, whether it is a phone line, cable line, or network cable can be plugged directly into the hub.

Software
Software proxies are much more economical than hardware proxies, and can support the same features as a dedicated hardware device. An example of a software proxy is Win98SE's Internet Connection Sharing. This is integrated into the Windows 98 Second Edition, and therefore free to use if you own that product. ICS supports basic firewall services, DHCP, DNS, and up to 254 users. There are other software products that can cache frequently used web pages to speed up the connection.

This is limited to 450 kbps, the standard speed of a serial port. It operates with a dual ended serial cable, and to each computer the connection operates as if it was a standard external dial-up modem. This is slow, and hardly ever used because of its slow transfer speed and 2 computer limit.

The most popular network technology because of its speed and flexibility. Often referred to as Unshielded Twisted Pair (UTP), this network uses cabling similar to telephone cable. Each cable is ended with RJ-45 connections, which look vary similar to phone jack connectors. The difference between RJ45 (UTP Ethernet) and RJ11 (telephone) connectors are that RJ45s are much larger because they have 8 pins instead of 4. The maximum distance between nodes or hubs is 185 meters.

Twisted pair networks are more flexible than BNC/coaxial Ethernet because it allows for the connection and re-connection of nodes without the need to shut down the entire network. And unlike coaxial Ethernet networks, instead of connecting nodes in series, each node has its own separate cable linked directly into a hub or switch. Hubs and switches duplicate and strengthen the incoming signal, and send it to its destination. Switches are more efficient because they send the signal to only the path which it needs to take, while hubs re-transmit the signal to all nodes.

10BaseT
This is exactly the same as 100BaseT, except for that is only able to transfer at speeds of up to 10MBps. It uses the same UTP cabling, but only requires the use of cable that category 3, instead of catgory 5. Category 3 cabling is the same cable that is used for telephones, and operates at speeds up to 10MHz.

Fast Ethernet 100baseT
Exactly the same as thin Ethernet, except that it operates at 100MHz, instead of 10MHz, allowing up to 100Mbps transfers. It also requires the use of category 5 cabling, which is know as Cat5. There are also, higher speed Ethernet connections that can transfer at speeds of 1 GBps and 10 Gbps, but they are usually very uncommon for use with computers other than servers.

Fast Ethernet 100baseTX
A type of Fast Ethernet that uses only 4 connections instead of 8. It still has the 2 twisted pairs, but no longer uses ground wires. Not very widely used because it can be less reliable than 100baseT.

This network design is often referred to as "cheapernet" because of its limited bandwidth and multiple design flaws. Thin Ethernet uses coaxial-cable to connect multiple computers in series. Each network card has a BNC connector which is used to allow the network to extend beyond it. Because coaxial cable is very sensitive to grounding problems, the chain of networked nodes has to be properly terminated at each end for it to work reliably. To do this, a BNC terminator has to be used on one end, and a BNC ground has to be used on the other. Thin Ethernet chain is limited to up to 30 nodes. Each thin Ethernet chain can be connected using a repeater and the total throughput is limited to 10Mbps transmission speed. Thin Ethernet doesn't require the use of hubs making it cheaper then UTP, but since it is unable to travel the distances that UTP is it usually isn't used that often.

Some newer network products allow a building's existing phone lines to be used as a network. It operates in the same way as DSL does, in that it uses the unused frequencies of the existing wires, allowing the phone to operate normally. The frequency used also, with most products, does not conflict with DSL frequencies. This technology currently allows up to 10Mbps transfers, which is comparable with 10baseT. The advantages of this is that no hub is required to connect multiple computers, and no extra cabling is required. Each computer on the network has to have an adapter that is plugged into a telephone jack. Distances should be nothing to worry about because they are able to operate at up to 185m.

Every house and computer uses power outlets, so this technology seems the most mobile. The technology is very similar to telephone line based networking, except wall outlets are used. The current products are limited to 500 kbps or less, and are able to operate at distances up to a quarter mile. Because the power lines for all houses in a city are connected, each packet must be encrypted so that other, unauthorized computers cannot capture it. The speed of this technology is limited because power lines suffer from sever fluctuations which can make signals unreliable. A possible advantage of power line and wireless technology over all of the others is its ability to travel between houses. If you want to set up a dedicated network with one of your neighbors, this is probably the best bet.

Wireless networking is a convenience because it offers hardly any restrictions to how computers have to be networked. Wireless networking communicate by high frequency radio waves in the 2 GHz to 6 GHz range. These high frequencies allow for the signals to pass-through most objects, such as walls, floors, ceilings, and across surprising distances. Depending on the quality of the networking product, computers can be up to 150 feet away indoors, or up to a mile apart outdoors. Like all non-hardwired connections, every packet is encrypted so that other networks will not be able to decode them. Transfer speeds are in the 5Mbps to 15Mbps range, and most devices attach to the computer via an ISA or PCI card.

The current USB standard is limited to 12Mbps, and so USB networking is as well. This allows for an easy 2 computer network, and operates similar to a null-modem cable. A dual ended USB cable is used, one end plugged into each computer. The benefits of this is that it is easy to install and requires no extra, and expensive hardware. The down sides of this is that it is limited to 2 computers and only an medium/slow transfer speed. This more a quick, temporary networking solution than it is a permanent one.

These 3 protocols are the main 3 used in all networks. All network cards are compatible and able to use these protocols, no matter what manufacture made them, or what they were designed for. A protocol is only a standard for communication, and is not depended on the type of hardware.

NetBEUI was developed by Microsoft as their propriety network solution for Windows based computers. This protocol was designed for use in a Local Area NetworK (LAN) setting. It uses the embedded number in the Network Interface Controller, called MAC addressing.

The MAC number is a 48-bit number; the first 24-bits are the manufacturer who made it, and the last 24-bits are the unique number for each card that the manufacturer produces. MAC numbers are reported as groups of 2 digits in hexadecimal form. An example of a MAC address is 00-60-67-67-E5-8A.

Some manufacturers have multiple 24 bit numbers in order to assure that each card is different. By use of this number, each card can be referenced without responses from other NIC's.

IPX/SPX was developed by Novell for use in their NOVELL-Netware servers. It is a fast protocol which can be routed. This protocol, like NetBEUI, also uses the embedded MAC number for addressing.

The IPX/SPX protocol was designed to be used with server running a file server product called "Netware". This server would use multiple network cards each joined to its own dedicated network, allowing one server to be used for multiple networks. The server would act as both a file server and a networking bridge. This multiple network approach was done to increase overall network performance.

IPX works in that the server would assign each separate network it was attached to a unique "NET" number. This NET number would make destinational routing a lot easier, all the server would have to do is check the packet's NET number to find the correct NET to route it to. This NET number was used in combination with the individual computer's MAC address to pinpoint the packet's destination quickly and efficiently. When no network server is connected to the network, computers must be manually configured with their own NET address.

Transmission Control Protocol
TCP/IP is a collection of protocols which was designed for use on a Wide Area Network (WAN) as opposed to a LAN. This protocol is physically slower than both IPX/SPX and NetBEUI, but allows for the connection of millions of computers. IP-address, subnet-mask, gateway, DNS, UTP, FTP, SMTP, DHCP, WINS, and TCP are all included in the TCP/IP suite.

Internet Protocol (IPv4, IPng)
TCP/IP uses IP's for addressing, rather than the NIC's embedded 48-bit number. An IP is a 32-bit number, usually expressed as 4 8-bit numbers in decimal form. For example, 24.245.65.3 would be a valid IP. To avoid confusion and multiple computers using the same IP, IP's are usually assigned by a server to each client using Dynamic Host Configuration Protocol (DHCP). This server keeps Another way to do it is for each computer to be manually assigned an IP, but this is only ever used on smaller networks where a DHCP server is unavailable.

IP has 32-bit addressing and allows for over 4 billion addresses. TCP/IP requires that no two IPs are the same, anywhere in the world. It also requires strict rules on how IPs are assigned. Because of poor planning and assigning problems, the number of unused IPs out of the 4 billion is very small. The internet works on routers and these routers connect high bandwidth cabling and direct packets in the direction that they have to travel. Routers analyze the first few bits of the IP to find the path that the packet has to travel. Because routers are physically incapable of remembering the destination path for every IP, IPs are grouped into similar paths. IPs have been divided into 4 separate classes.

Class	Header Bits								IP Range
	0	1	2	3	8	16	24	32
Class A	0	7-bit (128)				24-bit (16 777 216)			0.x.x.x to 127.x.x.x
Class B	1	0	14-bit (16 384)				16-bit (65 536)		128.x.x.x to 191.x.x.x
Class C	1	1	0	21-bit (2 097 152)				8-bit (256)	192.x.x.x to 223.x.x.x
Class D	1	1	1	0	28-bit (268 435 456)				224.x.x.x to 255.x.x.x
	Classid				Netid			Hostid

The first 3 classes, Class1, Class2, and Class3 are unicast transmissions, meaning that there is only one device in the entire world that is allowed to have that IP address. ClassD is a multicase IP address range, packets are sent to all addresses that have that netid. Because of limitations in the IPv4 protocol, this ClassD could never be practically applied.

The way that IP Classes are set up is that ClassA netid's are give to large corporations that have a need for a lot of IP's. This allows up to 128 different corporations to have over 16 million hostid addresses at their disposal. ClassB netid's are usually given to medium sized corporations, organizations or communities. This allows for over 16 thousand different organizations to have 65 thousand hostid addresses for their use. The over 2 million ClassC netids are what make up the bulk of the internet base. Each ClassC netid owner, such as an ISP or small network can have up to 256 hostid's attached to them. You can see how IP addresses are divided into groups. This allows routers to only have to know the correct path to and from each different netid. This makes their job a lot easier that having to remember the over 4 billion different addresses. Once the packet for a specific netid reaches that netid's gateway server, it is up to the gateway to route it properly from then on. But because there are some many ClassC addresses, and most companies who can't get ClassB netid's anymore, are going to need more than 256 IP addresses, Common Interdomain Routing Protocol (CIRP) was implemented. This allows multiple ClassC netid's to be combined in a miniature ClassB address. By grouping most ClassC addresses together, routers job a little bit easier.

The problem lies in the ClassA IP's. No organization has a need for 16 million hostid addresses. This means that a lot of IP's go unused, and there is no way to recover them. Routers send packets on a certain path based on their netid. If when the packet gets to the proper netid's gateway server and there is no hostid, the packet is discarded.

There are still many IP's which are reserved and cannot be used. These are addresses with 255 values, which are reserved for gateway functions, and the IP's in the range of 192.168.x.x, which are reserved for networks which are not connected to the internet, or use a proxy server for IP masquerading.

Internet Protocol Version 6 (IPv6)
IPv6 is the proposed IPv4 replacement. Because there are many downfalls of IPv4, and because useable IP addresses are rapidly running out because of rapid expansion of the internet, a solution has to be created. This solution is IPv6, and its major enhancement is the use of a 128-bit header instead of IPv4's 32-bit. IPv6 was designed to be backward compatible with IPv4, allowing for an easy, and gradual transition. IPv4 and IPv6 can be used concurrently on a signal system, and IPv6 can be tunneled through existing IPv4 networks. There are already IPv6 networks, and support for IPv6 in many different OSs.

3bits	m bits	n bits	o bits	p bits	125-(n+m+o+p)bits
010	ReglID	ProvdID	SubSCID	Subnet ID	IntfID

By structuring IP addresses by regional bounds, each region getting progressively smaller, routers at different stanges can compare different portions to find the right direction. Each cluster of bits gets more and more specific. This was designed to ease routers job while providing more than enough IP addresses for all future uses.

Along with new addressing, IP headers were reworked to provide more features.

IPv6 Header divided into 32-bits sections
4	8	12	16	20	24	28	32
Version	Priority	Flow Label
Payload Length				Next Header		# Of Hops
Source Address
Destination Address

Priority Information can be added here to mark the time importance of the packet. Network traffic such as streaming video, computer games, and user based response programs like Telnet have more importance placed on timely packet delivery than FTP and file transfer data. IPv6 leaves this header information to be used by programs that require a fast response, so that routers can give special priority to those packets over non-time critical ones.

Flow Labels This header information is to be used by future routers in speeding up and channeling data in the correct directions. If a stream of multiple packets is being sent, each will be marked by the flow label. This will allows supporting routers to quickly process and direct those packets because it is expecting successive packets to follow it.

Payload Length This field in the header is used to specify packet size, allowing packets to be potentially much bigger than IPv4 is capable of.

Next Header This is used to extend the capabilities of the header, to expand to more than the 128-bit limit. The usefulness in this lies with different encryption schemes, encodings and tunneling.

Hop Limit This is to be mostly used with multicast signals, to prevent the data from being recirculated after it has already reached its target group.

User Datagram Protocol
This is a replacement layer for TCP in TCP/IP that allows custom use of the data being encapsulated.

Domain Name System (DNS)
DNS is the simple conversion from host name to IP address. Every Internet connection needs to be able to access a DNS server if it wishes to make references based on host names rather than IP addresses. This is convenient, because instead of having to remember the IP address, of say Yahoo.com, you just have to remember Yahoo's name, which is Yahoo.com. DNS makes the Internet a much more user-friendly place.

All host names and IP addresses have to be registered at an organization called Internic. Internic keeps all of the information about who are using what IP addresses and what host names they correspond to. Internic keeps these name-IP lists on what are called public root servers. Whenever a host name has to be resolved, the request is usually sent to these servers.

If all DNS resolution requests were sent to these servers, the Internet would be really slow. There are thousands of names that have to be checked for every request which requires significant processing time. The solution to this was that each ISP has their own DNS server which records and caches these resolutions. Whenever a user of that ISP looks for a name, the request is sent to that ISP's DNS server. The DNS server checks its records first, and if the name is not found, it sends the request to the public root servers. DNS servers have what is called Time To Live (TTL) for each name. This means that each name is only cached for only a short period of time, depending on the DNS TTL settings. Most DNS servers operate with a TTL of around 2-3 days. This ensures that the requested name is accurate.