EthernetPrimer_Whitepaper

ETHERNET PRIMER

WHY ITS BETTER, FASTER, AND CHEAPER THAN OTHER NETWORK PROTOCOLS

By Wayne Rash

ETHERNET PRIMER - WHY ITS BETTER, FASTER, AND CHEAPER THAN OTHER NETWORK PROTOCOLS 2

Wayne Rash has been writing technical articles about computers and networking since the mid-1970s. He is a former columnist for Byte Magazine, a former Editor of InternetWeek, and currently performs technical reviews of networking, wireless, and data center products. He is the former Director of Network Integration for American Management Systems and is one of the founders of the Advanced Network Computing Laboratory at the University of Hawaii. He is based in Washington, DC and can be reached at [email protected].

Ethernet fills a number of roles in the enterprise far more than you might initially expect. This overview helps you understand the many ways in which the networking protocol is used in business, where it makes the most impact, and the effect CIOs should expect from upcoming innovations.

The once little-known networking protocol that began by providing a data link between islands in Hawaii is now, 40 years later, the overwhelmingly dominant means of communications throughout the world. Ethernet affects every type of communications, even though in most cases its nearly invisible.

Ethernet is the networking protocol thats behind virtually all local, metropolitan, and wide area networks. Because it is so broadly used, and because its based on a relatively simple protocol, Ethernet is easily managed, widely available, and broadly understood by the tech staff that run data centers. This in turn means that the Ethernet can cost less to implement, use, manage, and operate than any other type of networking. It also means that enterprises can use it more efficiently than other types of networking for their entire networking needs.

But because its so ubiquitous, the technology can be invisible to all but the most hard-core of technical staff. Its time for smaller businesses to get up to speed.

Ethernet Today

Youre an Ethernet user. Even if you dont have a computer in your office or home, and even if you never personally encounter a computer network, youre probably using Ethernet every day. The reason

is that Ethernet is at the heart of nearly all electronic communications, from your home telephone, to your cell phone and even to the car you drive. Somewhere along the way, the information you need travels over Ethernet.

Recent estimates put Ethernet usage at more than 85 percent in all networks worldwide, and at more than 95 percent for all local networks. The link that carries your telephone calls from one place to another almost certainly uses Ethernet. Likewise, the link between the cell phone tower and the telephone company is also likely to be Ethernet. And regardless of whether the portion of the Internet youre using at any given time is Ethernet (although it almost certainly is), the part of the Internet that reaches from the core of the Internet to your network, the so-called Last Mile, is Ethernet.

But Ethernet shows up in places other than the wires connecting your computers or wireless router to the cable system. Its also the backbone of nearly every data center, the connections between data centers, and the connections between business locations. Inside the data center, nearly every blade server and chassis has Ethernet on its backplane, and an Ethernet switch as one of the blades. In many cases, those same servers are connecting to storage using Ethernet either as a channel for iSCSI or Fibre Channel over Ethernet. Ethernet is, in short, ubiquitous.

In nearly every business, Ethernet requires a fairly standardized set of components that, when put together, create the company network. A similar thing happens on a smaller scale in small businesses and even in residential implementations, but the enterprise illuminates the core of how Ethernet is used.

Ethernet in Enterprise Computing

Since its commercial introduction in 1980, Ethernet has grown to be the de facto standard network protocol in enterprises at every level. What started out as an enterprise-class solution later found its way into a variety of other roles. What began as a means of driving high-speed printers quickly moved into high-speed (at the time) communications. Companies adopted Ethernet because it was readily available, relatively inexpensive,


and well understood at all levels in the IT department. Now with a penetration estimated above 95 percent for local networks, there are few organizations that arent using Ethernet as the only networking means.

While other technologies have moved from the shared resource model of networking, where everyone simultaneously shared a common networking cable (and available bandwidth), Ethernet remains in use. Today, Ethernet is almost always seen in a switched environment, where each network device has its own network segment resulting in an environment where collisions never happen. Bandwidth has grown by orders of magnitude to the point where Gigabit Ethernet is now standard issue for new computers.

Ethernet fills a number of roles in the enterprise. While all of it involves moving data, there are different types of data, and it is moved in different ways.

Ethernet on the Production LAN: This is what most people think about when they think about networking: tying together workstations, Wi-Fi access points, printers, scanners, and servers. Its also the pathway to the Internet and to the broader corporate Intranet.

The Enterprise Intranet: Most companies of any size have a central network that provides the backbone for its data traffic. This corporate backbone network is what ties together the Ethernet switches in various parts of the company and connects them to the data center, to remote parts of the enterprise, and in some companies to the cloud provider or to the companys backup and continuity sites. The enterprise intranet may also connect the enterprise to the campus network, to a metropolitan area network, or to a wide area network.

Wide Area Ethernet: While wide area networks have used a variety of protocols, the primary move currently is to Ethernet. Wide area Ethernet, which may be called Carrier Ethernet or Metro Ethernet, is designed to cover distances in excess of those normally found in the enterprise, but which are needed to provide communications between widely

A BRIEF HISTORY OF ETHERNET

Ethernet began as a means of wide-area communications when the University of Hawaii (UH) needed a way to provide access to a time-shared computer system located at the main campus on the island of Oahu by users on other islands. To accomplish this, the University staff created a radio system that would transmit data to all of the islands, and the stations needing to make the connection would transmit individually to the central station on Oahu.

Because several stations all used the same radio frequencies, the UH scientists designed what was then called AlohaNet so that the stations could listen for other stations already transmitting, and then wait a random amount of time before trying again. To do this, the data had to be sent in large enough chunks (called packets) that it was possible to tell that a packet was being transmitted.

Bob Metcalf, a PhD student working at UH at that time, wrote his dissertation on an advanced form of networking using the concepts of AlohaNet. Later, while working as a researcher at the Xerox Palo Alto Research Center (PARC), Metcalf wrote a memo describing what he then named Ethernet.

A few years later, the IEEE agreed to a series of technical standards for networking. One of those standards was Ethernet, which was given the designation IEEE 802.3. This standard included the original carrier-sensing techniques from AlohaNet along with collision detection (when two stations send data at the same time) as well as the physical layer (meaning the cables and electronics) standards. This set of standards has evolved as Ethernet networks serve a wider variety of purposes and carry different types of information at ever-increasing rates.

But originally, Ethernet had a single, almost prosaic purpose: to get data from mainframe computers to high-speed printers. It wasnt until shortly after that when computers started using Ethernet for communications. But those prosaic beginnings also meant that Ethernet was essentially a relatively simple method of communications. The protocol was defined by its ability to listen for potential problems, to (try to) avoid them, and then later to detect when data collisions (meaning, two computers sending data at the same time) happened, and then correct the problem.

That inherent simplicity was behind its success. Building Ethernet networks was relatively easy, relatively inexpensive, and as it turned out, highly flexible. While Ethernet originally had some competition from IBMs Token Ring protocol, that networking solution was limited in bandwidth, not particularly scalable, complex to implement, and it was troubled by hard-to-solve failure modes.


spaced areas. Examples of Metro Ethernet include networks that connect schools and school administration buildings throughout a county, very large industries or even the last-mile connectivity provided by a cable network. Carrier Ethernet is a related type of wide area Ethernet, meant for use by organizations such as phone companies; it provides a link between cell towers and central offices and connects central offices with each other.

Several types of Ethernet exist independently from the organizational structure in which they exist. These types are defined by the specific type of traffic that travels on the network, which in turn can affect the type of infrastructure that is chosen, and the management and operational characteristics. These uses can include:

Data Access: A data access network is what connects you to the Internet, the company server, or the network printer. Its also the type of network that connects the Wi-Fi hotspot to the rest of the company network.

Voice Services: Ethernet is an alternative to the traditional analog or plain old telephone service (POTS) voice network. Telephone switches have moved from analog PBX installations to PBX software that runs on a network server. Sometimes voice services are provided by cloud-based PBX software that runs on a server supporting many companies; these are reached using the company intranet, and from there connecting the Internet or a Metro or a private network to the cloud provider. A number of carriers provide these cloud- based or virtual PBX services, all of which require an Ethernet connection to each users phone.

Video Services: As Ethernet becomes more widely spread and as bandwidth becomes more readily available, centralized video storage has become highly cost-effective and highly practical. With Ethernets ability to accept prioritization of traffic, video and voice communications can traverse the network without being impeded.

Cloud Services: Widespread Ethernet has enabled organizations to centralize data storage, either as

a replacement for local storage, as a means of backup, or both. Cloud services make it possible for employees to access their data from anywhere, which in turn adds flexibility, while also reducing costs by consolidating several functions into a single network.

Data Center Ethernet

The way Ethernet is used in the data center is fundamentally different from how Ethernet is used elsewhere in an organization. The combination of relatively short distances coupled with extremely high bandwidth requirements mean that Ethernet exists in the data center in a way that would be unrecognizable to most managers. In fact, within the data center there are places where Ethernet more closely resembles cabinetry than it resembles a network.

While Data Center Ethernet is still Ethernet, it has a group of extensions that allow it to handle a variety of services that allow client to server, server to server and storage, and storage to storage communications that forms a unified communications fabric. It uses priority-based flow control and sends traffic over what are called virtual lanes that are essentially connections between input buffers and output queues. The network traffic combines data and storage area network traffic into a single fabric for greater efficiency and redundancy.

Data Center Ethernet has some enhancements that allow it to operate over multiple routes simultaneously to bond channels, and to work without cables. Its quite common for Data Center Ethernet to live on the backplanes of blade servers, and to interoperate with Ethernet fabric switches within each blade chassis.

The fact that its still Ethernet means that when the network emerges from the data center, it does so intact. Theres no need for protocol conversions required of other protocols, and no need to pass it through the switch or router that takes it from the data center to the next place its needed, whether its the enterprise network or to a Wide Area or Metropolitan Ethernet.


Metro Ethernet

Metropolitan Ethernet (MAN) is really just Ethernet that covers a wide area. But while its possible to operate a MAN using nothing but basic Ethernet, that isnt common. Normally, some other means of network infrastructure management helps support the network to improve reliability and management over wide areas.

The most visible Metro Ethernet networks are operated by cable companies. Those cable Internet products, marketed to everyone from consumers to businesses, become part of a Metro Ethernet network once they leave the building.

The two types of Metro Ethernet are point-to-point and multi-point to multi-point. Point-to-point Metro Ethernet operates much like other leased lines in that it simply connects one facility (a cell phone tower, for example) to another point (such as a telephone switching center) and it isnt shared. Multi-point to multi-point networks are more publicly visible and are the type of Metro Ethernet generally sold for Internet access, cloud services, and even private networking over shared infrastructure.

MANs can travel over a wide variety of networks and infrastructure including passive fiber, coaxial cable, and even fixed wireless. Tests conducted by the University of Hawaii and InternetWeek, a magazine formerly published by CMP media, successfully demonstrated microwave transmission of Ethernet on a MAN using multiprotocol label switching (MPLS) providing a near-gigabit bandwidth link while transmitting full-rate high-definition television images.

Metro Ethernet networks frequently use MPLS to simplify routing on large networks. The labeling also avoids complex network addresses, improves performance, and reduces complexity. The Ethernet content is encapsulated within the MPLS packets. Because MPLS ignores the content of the Ethernet packet during the time that its being transported across the network, its possible to mix different types of Ethernet traffic into a single network with little overhead.

ETHERNET IMPLEMENTATION FOR BUSINESS

Heres a look at what comprises almost every Ethernet implementation, although the details can vary widely depending on the business needs:

The Servers: The central server or servers create the basis for the network. They provide storage, delivery of centralized print, fax, and scanning, directory services, and applications. The servers may also contain the access to a larger storage pool such as a storage area network or to a private cloud.

Ethernet Switches: In most networks, there are two types of Ethernet switches. Edge switches are the devices that accept physical connections from client workstations, Wi-Fi access points, printers, scanners, and other networked equipment, and send it along to where it needs to go. Most network switches read the information contained in the Ethernet data so that they know where the data is coming from and the destination of the data. The switch then creates a connection between the two. Core switches accept connections from the edge switches and send it to other parts of the network, such as other edge switches or to the central servers.

Ethernet Routers: A router operates in a manner similar to a switch, but normally only sends data between the internal network and the outside world, or between two separate parts of the internal network. The device that allows a company to communicate from its internal network to the outside world using the Internet is a router.

Firewalls: These are devices that prevent unwanted network traffic from entering the internal network. Sometimes firewalls are part of the router that leads to the outside world and the Internet, but in a corporate setting they are more often stand-alone devices. Their function is to inspect traffic coming into the network for malware (such as worms and viruses), unauthorized access attempts, and other kinds of attacks. They may also inspect outgoing traffic for specific types of data, such as protected information. Firewalls usually keep detailed electronic logs, which permit later analysis when theres a reason to believe information may have escaped from the local network. Some enterprises also use firewalls to separate sections of the network that should not have easy access to each other, such as protecting the accounting and payroll departments from access by the company employees who arent part of those departments.

Network Interfaces: Everything that attaches to the network has a network interface, which allows the computer or other device to send data over the network. (Otherwise its like playing music on a stereo without speakers or other type of sound output.) This component may be installed in a computer or it may be built in to the computer or other network device.

(Continued on following page)


In some cases, such as when MPLS passes through routing centers, the packets may be encapsulated within Ethernet. That leads to a situation where for a time the network traffic is Ethernet contained within MPLS packets, which in turn are contained in Ethernet packets. When the Ethernet packets (regardless of how theyre encapsulated) emerge at the end of the MPLS network, they revert to pure Ethernet again. The process of handling them returns to the simplicity of Ethernet, and uses standard Ethernet equipment.

The key to Metropolitan Ethernet is the fact that its still Ethernet. What happens in the middle those places where MPLS plays a role doesnt matter to end users at any level. What does matter is that when the MAN is connected to the network inside the enterprise, its back to being pure Ethernet. This means that the standard and reasonably priced equipment that you need to handle your network is the equipment the IT staff is used to working with.

The process of taking the Ethernet packets and introducing them into the Metro Ethernet may or may not take place on premises. Normally, the network provider installs a MAN interface somewhere on the downstream end of the enterprise network. Depending on the type of provider, your company size, and its anticipated bandwidth requirements, this interface can be anything from a cable modem (which is, by the way, a misnomer since its not actually a modem), to a fiber interface module, to a rack-sized equipment module that is installed in the data center.

The network provider has control over certain aspects of the network interface, including the ability to provision it for specific types of Ethernet traffic and to provision it for varying levels of network bandwidth. What this means is that the network provider can remotely increase the available bandwidth for special needs without having to send out a technician. For example, if an e-commerce company is planning a big sale and as a result is expecting more than the normal amount of network traffic, the provider can turn up the bandwidth to accommodate the situation.

ETHERNET IMPLEMENTATION FOR BUSINESS (CONT.)

Network Media: The network media you probably think of first are those Ethernet cables that plug into the back of a computer and into a wall jack. They look like telephone cables and are similar in construction, but they are not interchangeable. Other network cables lead from the wall jack to the edge router, which is connected to the other network devices, and to the backbone network. Up to this point, these network cables are usually copper, just like telephone cables. But when the cables leave the switch, they may be optical fiber.

- Copper network cables are graded by the bandwidth they need to support and the distance they need to traverse before they reach the switch. Most (relatively) short copper cables are called Category 5, which support Gigabit Ethernet over fairly short distances. Category 5E is more reliable over somewhat longer distances, and Category 6 is for even more reliable high-bandwidth needs over longer distances.

- Fiber network cables are used by Ethernet switches and other components where reliability is crucial. Optical fiber is immune from electrical interference, its extremely durable, and it can be used over extremely long distances. Most enterprises use optical fiber for their backbone networks because of its reliability. However, optical fiber is more expensive than copper, and the network interfaces are more expensive. Because of the expense, optical fiber interfaces are rarely found outside of the servers, the data center, and other central equipment. Some routers and firewalls include optical fiber interfaces. Workstations and wireless access points almost never include them.

- Radio: Network connections that dont require long distances or large amounts of bandwidth frequently use radio, commonly known as Wi-Fi, as the means of connecting user workstations with the network. Wi-Fi is a variation of Ethernet that allows more flexibility in the physical location of the workstation or other network device, but its more prone to interference than is wired Ethernet. Wireless networking depends on an access point that communicates with the devices using Wi-Fi and the Ethernet network. Implemented properly, Wi-Fi is secure enough for most uses, and is reliable. However, Wi-Fi is a shared resource, meaning that its possible for others who should not have access to intercept signals, so encryption of Wi-Fi signals is essential. In smaller implementations, Wi-Fi may be combined with the Ethernet router and firewalls, but this configuration is very rare in the enterprise.

Intrusion Detection and Prevention Devices: While not essential to the existence of an Ethernet network, these security devices watch network activity for suspicious events. These events could include an unauthorized user who somehow got access to the network, to the installation of an unauthorized wireless router, or to network traffic containing prohibited material. While not required to make Ethernet work, these devices are essential to the security of the network.

Whats interesting is that everything on the list of network equipment (with the exception of the security devices) has been part of Ethernet in one way or another since its inception.


But of course theres a limit, in that the network provider can only provide the capacity that the infrastructure is capable of delivering. Beyond that, the network provider needs to upgrade the infrastructure or add additional infrastructure. This is the time when a technician may need to visit, but the result means that the enterprise can meet its capacity needs as its business expands.

The Future of Ethernet

The limits to the capacity of Ethernet are few, and are mostly governed by the laws of physics. An Ethernet data packet, for example, can only travel over a copper wire at the speed of light in copper roughly half the speed of light in a vacuum, which is the speed you learned about in school. The speed of Ethernet packets in fiber is faster than in copper, but still limited depending on the fibers type and composition. This is important in two areas. First, theres a minimum physical length to an Ethernet cable because the data packet has to leave its source before it arrives at its destination. Theres also a maximum length, but thats more related to the Ethernet infrastructure electronics and their timing requirements. Those timing requirements are set when each Ethernet standard is set.

And in reality, its the standards-setting process that places the greatest limits on Ethernet bandwidth. The original Ethernet specification provided a bandwidth of 10 megabits per second, and the standard was written so that every provider of Ethernet equipment could interoperate. Likewise, as Ethernet capacity grew with improvements in electronics, new standards supported 100, then 1,000 megabits per second.

Gigabit Ethernet 1,000 megabits per second, or mbps is the current standard for most enterprise networking. Virtually every networking device and interface available supports Gigabit Ethernet, including everything from consumer laptops to servers in the data center. But larger data users need greater capacity. 10 Gigabit Ethernet is widely used in metropolitan networks, within data centers, and in some cases as enterprise backbones.

40 and 100 Gigabit Ethernet hardware is already available from a few vendors. The current push is for 400

Gigabit per second Ethernet, which isnt yet available, but is already the subject of an IEEE standards effort. A few vendors have already begun work on experimental 400 gbps equipment.

The next obvious step is Terabit Ethernet, which for several years was dismissed as being too fast to be useful. But of course needs change, and with the explosive growth of cloud computing, the general realization is that theres no such thing as being too fast. As a result, teams of researchers have begun work to create the first Terabit Ethernet. All of this race to the theoretical top may seem like the network version of the old saying from New York media circles, You cant be too rich or too thin. Except that in the case of networks, you cant be too fast.

Ethernet and Your Company

The fact that researchers are finding ways to increase the capacity of Ethernet to ever more breathtaking levels is good news for any organization that depends on cloud services or even has a plan for increasing its e-commerce presence. The network capacity that providers can provision is unlikely to ever become a bottleneck. It will, of course, take a while for 400 Gigabit or Terabit Ethernet to become widely available once theyre developed and their standards are approved.

Fortunately, one major goal of the standards bodies is to make it possible for this higher capacity networking to operate within existing physical infrastructure. This means that when 10 Gigabit Ethernet over copper becomes widely available, it should be capable of functioning on most existing copper infrastructure, as long as that cabling meets Category 6 (for short distances) or Category 6E or 7 for distances up to 100 meters. Anything more than that will require fiber, but most existing fiber infrastructure will likely work. Ethernet with data rates faster than 10 Gigabit Ethernet are currently aimed at fiber infrastructure and exist in the data center or the metropolitan network arena.

While 10 Gigabit Ethernet is readily available, currently its only used for switch-to-switch communications, or for switch-to-server communications within data centers. When 40 and 100 Gigabit Ethernet becomes commonly


available, their likely use will be in data centers and wide and metro area uses. The chances of seeing Terabit Ethernet in your enterprise are very slim, unless your enterprise is a Fortune 50 company.

What this means to the enterprise is that its safe to plan on better access to cloud services, remote locations, and off-site fail-over facilities. But it also means that theres little if any risk in building out your in-house infrastructure with todays Ethernet standards in mind. Most of the existing infrastructure is scalable, and the part thats not (such as the servers, switches, routers, and the like) will reach the end of its economic life before its technical limitations become an issue. But its important to keep in mind that changes will happen eventually, so that when data access needs to grow, the Ethernet infrastructure will be there to support it.

Conclusion

In the 40 years since Ethernet was invented, this protocol has grown to the point that its a real contender for networking except in very specialized circumstances. Even industrial uses that once used specialized protocols for applications such as factory floor automation have moved to Ethernet, or to Wi-Fi connected to Ethernet.

The overwhelming use of Ethernet for enterprise use, and the very broad use of Ethernet in one form or another for Wide Area and Metropolitan Area Networks, has changed the way organizations use their remote assets and their access to data. Ethernet, which is widely understood in the data center and the IT shop, now is a significant standard reaching from the workstation to the cloud. In addition to simplifying the process of networking, this service can lower the price of the infrastructure, lower the cost of operation, and help ensure that staff is available that is familiar with the network and the equipment that supports it.

The result of this standardization is that the total cost of ownership of an Ethernet infrastructure can be much lower than with any other networking solution. In addition, because of its ubiquitous nature, Ethernet is easy and fast to implement. Because of this, Ethernet is rapidly approaching the point where its no longer a collection of products, cables, and software. Ethernet

has become a widely available service in which the network quietly fades into the background, just as other utilities such as electrical wiring and plumbing. Ultimately, Ethernet will continue to exist, of course, but it will fade from view to most, because it will simply exist as ubiquitous connectivity that provides whatever level of service is needed at the time.

ENTERPRISESOLUTIONSWHY CHOOSE COMCAST FOR YOUR ORGANIZATION? Services highly scalable, secure

and reliable

Network Americas leading enhanced network for Internet, Phone and TV

Value Comcast is the clear choice just compare us to the competition

Support 24/7 customer support from highly experienced local account teams

For more information, visit business.comcast.com

SPEAK WITH A COMCAST

BUSINESS REPRESENTATIVE

(877) 542-3963

EthernetPrimer_Whitepaper

Documents

ethernet usage

ethernet todayyoure

ethernet user

ethernet primerwhy

type of networking

types of networking

networking protocol

computer network