Networking

10-Gigabit Ethernet, sometimes called 10-GbE or 10 GigE, is the latest improvement on the Ethernet standard, ratified in 2003 for fiber as the 802.3ae standard, in 2004 for twinax cable as the 802.3ak standard, and in 2006 for UTP as the 802.3an standard.

10-Gigabit Ethernet offers ten times the speed of Gigabit Ethernet. This extraordinary throughput plus compatibility with existing Ethernet standards has resulted in 10-Gigabit Ethernet quickly becoming the new standard for high-speed network backbones, largely supplanting older technologies such as ATM over SONET. 10-Gigabit Ethernet has even made inroads in the area of storage area networks (SAN) where Fibre Channel has long been the dominant standard. This new Ethernet standard offers a fast, simple, relatively inexpensive way to incorporate super high-speed links into your network.

Because 10-Gigabit Ethernet is simply an extension of the existing Ethernet standards family, it’s a true Ethernet standard—it’s totally backwards compatible and retains full compatibility with 10-/100-/1000-Mbps Ethernet. It has no impact on existing Ethernet nodes, enabling you to seamlessly upgrade your network with straightforward upgrade paths and scalability.

10-Gigabit Ethernet is less costly to install than older high-speed standards such as ATM. And not only is it relatively inexpensive to install, but the cost of network maintenance and management also stays low—10-Gigabit Ethernet can easily be managed by local network administrators.

10-Gigabit Ethernet is also more efficient than other high-speed standards. Because it uses the same Ethernet frames as earlier Ethernet standards, it can be integrated into your network using switches rather than routers. Packets don’t need to be fragmented, reassembled, or translated for data to get through.

Unlike earlier Ethernet standards, which operate in half- or full-duplex, 10-Gigabit Ethernet operates in full-duplex only, eliminating collisions and abandoning the CSMA/CD protocol used to negotiate half-duplex links. It maintains MAC frame compatibility with earlier Ethernet standards with 64- to 1518-byte frame lengths. The 10-Gigabit standard does not support jumbo frames, although there are proprietary methods for accommodating them.

Fiber 10-Gigabit Ethernet standards
There are two groups of physical-layer (PHY) 10-Gigabit Ethernet standards for fiber: LAN-PHY and WAN-PHY.

LAN-PHY is the most common group of standards. It’s used for simple switch and router connections over privately owned fiber and uses a line rate of 10.3125 Gbps with 64B/66B encoding.

The other group of 10-Gigabit Ethernet standards, WAN-PHY, is used with SONET/SDH interfaces for wide area networking across cities, states—even internationally.

LAN-PHY
10GBASE-SR (Short-Range) is a serial short-range fiber standard that operates over two multimode fibers. It has a range of 26 to 82 meters (85 to 269 ft.) over legacy 62.5-µm 850-nm fiber and up to 300 meters (984 ft.) over 50-µm 850-nm fiber.

10GBASE-LR (Long-Range) is a serial long-range 10-Gbps Ethernet standard that operates at ranges of up to 25 kilometers (15.5 mi.) on two 1310-nm single-mode fibers.

10GBASE-ER (Extended-Range) is similar to 10GBASE-LR but supports distances up to 40 kilometers (24.9 mi.) over two 1550-nm single-mode fibers.

10GBASE-LX4 uses Coarse-Wavelength Division Multiplexing (CWDM) to achieve ranges of 300 meters (984 ft.) over two legacy 850-nm multimode fibers or up to 10 kilometers (6.2 mi.) over two 1310-nm single-mode fibers. This standard multiplexes four data streams over four different wavelengths in the range of 1300 nm. Each wavelength carries 3.125 Gbps to achieve 10-Gigabit speed.

WAN-PHY
In fiber-based Gigabit Ethernet, the 10GBASE-SR, 10GBASE-LR, and 10GBASE-ER LAN-PHY standards have WAN-PHY equivalents called 10GBASE-SW, 10GBASE-LW, and 10GBASE-EW. There is no WAN-PHY standard corresponding to 10GBASE-LX4.

WAN-PHY standards are designed to operate across high-speed systems such as SONET and SDH. These systems are often telco operated and can be used to provide high-speed data delivery worldwide. WAN-PHY 10-Gigabit Ethernet operates within SDH and SONET using an SDH/SONET frame running at 9.953 Gbps without the need to directly map Ethernet frames into SDH/SONET.

WAN-PHY is transparent to data—from the user’s perspective it looks exactly the same as LAN-PHY.

10-Gigabit Ethernet over Copper
10GBASE-CX4
10GBASE-CX4 is a standard that enables Ethernet to run over CX4 cable, which consists of four twinaxial copper pairs bundled into a single cable. CX4 cable is also used in high-speed InfiniBand® and Fibre Channel storage applications. Although CX4 cable is somewhat less expensive to install than fiber optic cable, it’s limited to distances of up to 15 meters. Because this standard uses such a specialized cable at short distances, 10GBASE-CX4 is generally used only in limited data center applications such as connecting servers or switches.

10GBASE-Kx
10GBASE-Kx is backplane 10-Gigabit Ethernet and consists of two standards. 10GBASE-KR is a serial standard compatible with 10GBASE-SR, 10GBASE-LR, and 10GBASE-ER. 10GBASE-KX4 is compatible with 10GBASE-LX4. These standards use up to 40 inches of copper printed circuit board with two connectors in place of cable. These very specialized standards are used primarily for switches, routers, and blade servers in data center applications.

10GBASE-T
10GBASE-T is the 10-Gigabit standard that uses the familiar shielded or unshielded copper UTP cable. It operates at distances of up to 55 meters (180 ft.) over existing Category 6 cabling or up to 100 meters (328 ft.) over augmented Category 6, or “6a,” cable, which is specially designed to reduce crosstalk between UTP cables. Category 6a cable is somewhat bulkier than Category 6 cable but retains the familiar RJ-45 connectors.

To send data at these extremely high speeds across four-pair UTP cable, 10GBASE-T uses sophisticated digital signal processing to suppress crosstalk between pairs and to remove signal reflections.

10-Gigabit Ethernet Applications
> 10-Gigabit Ethernet is already being deployed in applications requiring extremely high bandwidth:
> As a lower-cost alternative to Fibre Channel in storage area networking (SAN) applications.
> High-speed server interconnects in server clusters.
> Aggregation of Gigabit segments into 10-Gigabit Ethernet trunk lines.
> High-speed switch-to-switch links in data centers.
> Extremely long-distance Ethernet links over public SONET infrastructure.

Although 10-Gigabit Ethernet is currently being implemented only by extremely high-volume users such as enterprise networks, universities, telecommunications carriers, and Internet service providers, it’s probably only a matter of time before it’s delivering video to your desktop. Remember that only a few years ago, a mere 100-Mbps was impressive enough to be called “Fast Ethernet.”

As workstations and servers migrated from ordinary 10-Mbps Ethernet to 100-Mbps speeds, it became clear that even greater speeds were needed. Gigabit Ethernet was developed for an even faster Ethernet standard to handle the network traffic generated on the server and backbone level by Fast Ethernet. Gigabit Ethernet delivers an incredible 1000 Mbps (or 1 Gbps), 100 times faster than 10BASE-T. At that speed, Gigabit Ethernet can handle even the traffic generated by campus network backbones. Plus it provides a smooth upgrade path from 10-Mbps Ethernet and 100-Mbps Fast Ethernet at a reasonable cost.

Compatibility
Gigabit Ethernet is a true Ethernet standard. Because it uses the same frame formats and flow control as earlier Ethernet versions, networks readily recognize it, and it’s compatible with older Ethernet standards. Other high-speed technologies (ATM, for instance) present compatibility problems such as different frame formats or different hardware requirements.

The primary difference between Gigabit Ethernet and earlier implementations of Ethernet is that Gigabit Ethernet almost always runs in full-duplex mode, rather than the half-duplex mode commonly found in 10- and 100-Mbps Ethernet.

One significant feature of Gigabit Ethernet is the improvement to the Carrier Sense Multiple Access with Collision Detection (CSMA/CD) function. In half-duplex mode, all Ethernet speeds use the CSMA/CD access method to resolve contention for shared media. For Gigabit Ethernet, CSMA/CD has been enhanced to maintain the 200-meter (656.1-ft.) collision diameter.

Affordability and adaptability
You can incorporate Gigabit Ethernet into any standard Ethernet network at a reasonable cost without having to invest in additional training, cabling, management tools, or end stations. Because Gigabit Ethernet blends so well with your other Ethernet applications, you have the flexibility to give each Ethernet segment exactly as much speed as it needs—and if your needs change, Ethernet is easily adaptable to new network requirements.

Gigabit Ethernet is the ideal high-speed technology to use between 10-/100-Mbps Ethernet switches or for connection to high-speed servers with the assurance of total compatibility with your Ethernet network.

When Gigabit Ethernet first appeared, fiber was crucial to running Gigabit Ethernet effectively. Since then, the IEEE802.3ab standard for Gigabit over Category 5 cable has been approved, enabling short stretches of Gigabit speed over existing copper cable. Today, you have many choices when implementing Gigabit Ethernet:

1000BASE-X
1000BASE-X refers collectively to the IEEE802.3z standards: 1000BASE-SX, 1000BASE-LX, and 1000BASE-CX.

1000BASE-SX
The “S“ in 1000BASE-SX stands for “short.“ It uses short wavelength lasers, operating in the 770- to 860-nanometer range, to transmit data over multimode fiber. It’s less expensive than 1000BASE-LX, but has a much shorter range of 220 meters over typical 62.5-µm multimode cable.

1000BASE-LX
The “L“ stands for “long.“ It uses long wavelength lasers operating in the wavelength range of 1270 to 1355 nanometers to transmit data over single-mode fiber optic cable. 1000BASE-LX supports up to 550 meters over multimode fiber or up to 10 kilometers over single-mode fiber.

1000BASE-CX
The “C“ stands for “copper.“ It operates over special twinax cable at distances of up to 25 meters. This standard never really caught on.

Gigabit over CAT5—1000BASE-TX
The 802.3ab specification, or 1000BASE-TX, enables you to run IEEE-compliant Gigabit Ethernet over copper twisted-pair cable at distances of up to 100 meters of CAT5 or higher cable.

Gigabit Ethernet uses all four twisted pairs within the cable, unlike 10BASE-T and 100BASE-TX, which only use two of the four pairs. It works by transmitting 250 Mbps over each of the four pairs in 4-pair cable.

Media converters interconnect different cable types such as twisted pair, fiber, and coax within an existing network. They are often used to connect newer Ethernet equipment to legacy cabling. They can also be used in pairs to insert a fiber segment into copper networks to increase cabling distances and enhance immunity to electromagnetic interference (EMI).

Traditional media converters are purely Layer 1 devices that only convert electrical signals and physical media. They don’t do anything to the data coming through the link so they’re totally transparent to data. These converters have two ports—one port for each media type. Layer 1 media converters only operate at one speed and cannot, for instance, support both 10-Mbps and 100-Mbps Ethernet.

Some media converters are more advanced Layer 2 Ethernet devices that, like traditional media converters, provide Layer 1 electrical and physical conversion. But, unlike traditional media converters, they also provide Layer 2 services—in other words, they’re really switches. This kind of media converter often has more than two ports, enabling you to, for instance, extend two or more copper links across a single fiber link. They also often feature autosensing ports on the copper side, making them useful for linking segments operating at different speeds.

Media converters are available in standalone models that convert between two different media types and in chassis-based models that connect many different media types in a single housing.

Rent an apartment

Standalone converters convert between two media. But, like a small apartment, they can be outgrown. Consider your current and future applications before selecting a media converter. Standalone converters are available in many configurations, including 10BASE-T to multimode or single-mode fiber, 10BASE-T to Thin coax (ThinNet), 10BASE-T to thick coax (standard Ethernet), CDDI to FDDI, and Thin coax to fiber. 100BASE-T and 100BASE-FX models that connect UTP to single- or multimode fiber are also available. With the development of Gigabit Ethernet (1000 Mbps), media converters have been created to make the transition to high-speed networks easier.

...or buy a house.

Chassis-based or modular media converters are normally rackmountable and have slots that house media converter modules. Like a well-planned house, the chassis gives you room to grow. These are used when many Ethernet segments of different media types need to be connected in a central location. Modules are available for the same conversions performed by the standalone converters, and 10BASE-T, 100BASE-TX, 100BASE-FX, and Gigabit modules may also be mixed.

Media converters transparently convert the incoming electrical signal from one cable type and then transmit it over another type—thick coax to Thin, UTP to fiber, and so on. Traditionally, media converters were purely Layer 1 devices that only converted electrical signals and physical media and didn’t do anything to the data coming through the link.

Today’s media converters, however, are often more advanced Layer 2 Ethernet devices that, like traditional media converters, provide Layer 1 electrical and physical conversion. But, unlike traditional media converters, they also provide Layer 2 services and route Ethernet packets based on MAC address. These media converters are often called media converter switches, switching media converters, or Layer 2 media converters. They enable you to have multiple connections rather than just one simple in-and-out connection. And because they’re switches, they increase network efficiency.

Media converters are often used to connect newer 100-Mbps, Gigabit Ethernet, or ATM equipment to existing networks, which are generally 10BASE-T, 100BASE-T, or a mixture of both. They can also be used in pairs to insert a fiber segment into copper networks to increase cabling distances and enhance immunity to electromagnetic interference.

Rent an apartment…
Media converters are available in standalone models that convert between two different media types and in chassis-based models that house many media converters in a a single chassis.

Standalone models convert between two media. But, like a small apartment, they can be outgrown.

Consider your current and future applications before selecting a media converter. A good way to anticipate future network requirements is to choose media converters that work as standalone devices but can be rackmounted if needed later.

…or buy a house.
Chassis-based or modular media converter systems are normally rackmountable and have slots to house media converter modules. Like a well-planned house, the chassis gives you room to grow. These are used when many Ethernet segments of different media types need to be connected in a central location. Modules are available for the same conversions performed by the standalone converters, and they enable you to mix different media types such as 10BASE-T, 100BASE-TX, 100BASE-FX, ATM, and Gigabit modules. Although enterprise-level chassis-based systems generally have modules that can only be used in a chassis, many midrange systems feature modules that can be used individually or in a chassis.

Fiber versus copper.

When planning a new or upgraded cabling infrastructure, you have two basic choices: fiber or copper. Both offer superior data transmission. The decision on which one to use may be difficult. It will often depend on your current network, your future networking needs, and your particular application, including bandwidth, distances, environment, cost, and more. In some cases, copper may be a better choice; in other situations, fiber offers advantages.

Although copper cable is currently more popular and much more predominant in structured cabling systems and networks, fiber is quickly gaining fans.

Fiber optic cable is becoming one of the fastest-growing transmission mediums for both new cabling installations and upgrades, including backbone, horizontal, and even desktop applications. Fiber optic cable is favored for applications that need high bandwidth, long distances, and complete immunity to electrical interference. It’s ideal for high data-rate systems such as Gigabit Ethernet, FDDI, multimedia, ATM, SONET, Fibre Channel, or any other network that requires the transfer of large, bandwidth-consuming data files, particularly over long distances. A common application for fiber optic cable is as a network backbone, where huge amounts of data are transmitted. To help you decide if fiber is right for your new network or if you want to migrate to fiber, take a look at the following:

The advantages of fiber.

Greater bandwidth-Because fiber provides far greater bandwidth than copper and has proven performance at rates up to 10 Gbps, it gives network designers future-proofing capabilities as network speeds and requirements increase. Also, fiber optic cable can carry more information with greater fidelity than copper wire. That’s why the telephone networks use fiber, and many CATV companies are converting to fiber.

Low attenuation and greater distance-Because the fiber optic signal is made of light, very little signal loss occurs during transmission so data can move at higher speeds and greater distances. Fiber does not have the 100-meter (304.8-ft.) distance limitation of unshielded twisted-pair copper (without a booster). Fiber distances can range from 300 meters to 40 kilometers, depending on the style of cable, wavelength, and network. (Fiber distances are typically measured in metric units.) Because fiber signals need less boosting than copper ones do, the cable performs better.

Fiber networks also enable you to put all your electronics and hardware in one central location, instead of having wiring closets with equipment throughout the building.

Security-Your data is safe with fiber cable. It does not radiate signals and is extremely difficult to tap. If the cable is tapped, it’s very easy to monitor because the cable leaks light, causing the entire system to fail. If an attempt is made to break the security of your fiber system, you’ll know it.

Immunity and reliability-Fiber provides extremely reliable data transmission. It’s completely immune to many environmental factors that affect copper cable. The fiber is made of glass, which is an insulator, so no electric current can flow through. It is immune to electromagnetic interference and radio-frequency interference (EMI/RFI), crosstalk, impedance problems, and more. You can run fiber cable next to industrial equipment without worry. Fiber is also less susceptible to temperature fluctuations than copper is and can be submerged in water.

Design-Fiber is lightweight, thin, and more durable than copper cable. And, contrary to what you might think, fiber optic cable has pulling specifications that are up to ten times greater than copper cable’s. Its small size makes it easier to handle, and it takes up much less space in cabling ducts. Although fiber is still more difficult to terminate than copper is, advancements in connectors are making temination easier. In addition, fiber is actually easier to test than copper cable.

Migration-The proliferation and lower costs of media converters are making copper to fiber migration much easier. The converters provide seamless links and enable the use of existing hardware. Fiber can be incorporated into networks in planned upgrades.

Standards-New TIA/EIA standards are bringing fiber closer to the desktop. TIA/EIA-785, ratified in 2001, provides a cost-effective migration path from 10-Mbps Ethernet to 100-Mbps Fast Ethernet over fiber (100BASE-SX). A recent addendum to the standard eliminates limitations in transceiver designs. In addition, in June 2002, the IEEE approved a 10-Gigabit Ethernet standard.

Costs-The cost for fiber cable, components, and hardware is steadily decreasing. Installation costs for fiber are higher than copper because of the skill needed for terminations. Overall, fiber is more expensive than copper in the short run, but it may actually be less expensive in the long run. Fiber typically costs less to maintain, has much less downtime, and requires less networking hardware. And fiber eliminates the need to recable for higher network performance.

Multimode or single-mode, duplex or simplex?

Multimode-Multimode fiber optic cable can be used for most general fiber applications. Use multimode fiber for bringing fiber to the desktop, for adding segments to your existing network, or in smaller applications such as alarm systems. Multimode cable comes with two different core sizes: 50 micron or 62.5 micron.

Single-mode-Single-mode is used over distances longer than a few miles. Telcos use it for connections between switching offices. Single-mode cable features an 8.5-micron glass core.

Duplex-Use duplex multimode or single-mode fiber optic cable for applications that require simultaneous, bidirectional data transfer. Workstations, fiber switches and servers, fiber modems, and similar hardware require duplex cable. Duplex is available in single- and multimode.

Simplex-Because simplex fiber optic cable consists of only one fiber link, you should use it for applications that only require one-way data transfer. For instance, an interstate trucking scale that sends the weight of the truck to a monitoring station or an oil line monitor that sends data about oil flow to a central location. Simplex fiber comes in single- and multimode types.

50- vs. 62.5-micron cable.

Although 50-micron fiber cable features a smaller core, which is the light-carrying portion of the fiber, both 62.5- and 50-micron cable feature the same glass cladding diameter of 125 microns. You can use both in the same types of networks, although 50-micron cable is recommended for premise applications: backbone, horizontal, and intrabuilding connections, and should be considered especially for any new construction and installations. And both can use either LED or laser light sources.

The big difference between 50-micron and 62.5-micron cable is in bandwidth-50-micron cable features three times the bandwidth of standard 62.5-micron cable, particularly at 850 nm. The 850-nm wavelength is becoming more important as lasers are being used more frequently as a light source.

Other differences are distance and speed. 50-micron cable provides longer link lengths and/or higher speeds in the 850-nm wavelength. See the table below.

The ferrules: ceramic or composite?

As a general rule, use ceramic ferrules for critical network connections such as backbone cables or for connections that will be changed frequently, like those in wiring closets. Ceramic ferrules are more precisely molded and fit closer to the fiber, which gives the fiber optic cables a lower optical loss.

Use composite ferrules for connections that are less critical to the network’s overall operation and less frequently changed. Like their ceramic counterparts, composite ferrules are characterized by low loss, good quality, and a long life. However, they are not as precisely molded and slightly easier to damage, so they aren’t as well-suited for critical connections.

Testing and certifying fiber optic cable.

If you’re accustomed to certifying copper cable, you’ll be pleasantly surprised at how easy it is to certify fiber optic cable because it’s immune to electrical interference. You only need to check a few measurements.

Attenuation (or decibel loss)-Measured in decibels/kilometer (dB/km), this is the decrease of signal strength as it travels through the fiber cable. Generally, attenuation problems are more common on multimode fiber optic cables.

Return loss-This is the amount of light reflected from the far end of the cable back to the source. The lower the number, the better. For example, a reading of -60 decibels is better than -20 decibels. Like attenuation, return loss is usually greater with multimode cable.

Graded refractive index-This measures how the light is sent down the fiber. This is commonly measured at wavelengths of 850 and 1300 nanometers. Compared to other operating frequencies, these two ranges yield the lowest intrinsic power loss. (NOTE: This is valid for multimode fiber only.)

Propagation delay-This is the time it takes a signal to travel from one point to another over a transmission channel.

Optical time-domain reflectometry (OTDR)-This enables you to isolate cable faults by transmitting high-frequency pulses onto a cable and examining their reflections along the cable. With OTDR, you can also determine the length of a fiber optic cable because the OTDR value includes the distance the optic signal travels.

There are many fiber optic testers on the market today. Basic fiber optic testers function by shining a light down one end of the cable. At the other end, there’s a receiver calibrated to the strength of the light source. With this test, you can measure how much light is going to the other end of the cable. Generally, these testers give you the results in dB lost, which you then compare to the loss budget. If the measured loss is less than the number calculated by your loss budget, your installation is good.

Newer fiber optic testers have a broad range of capabilities. They can test both 850- and 1300-nanometer signals at the same time and can even check your cable for compliance with specific standards.

Fiber precautions.

A few properties particular to fiber optic cable can cause problems if you aren’t careful during installation.

Intrinsic power loss-As the optic signal travels through the fiber core, the signal inevitably loses some speed through absorption, reflection, and scattering. This problem is easy to manage by making sure your splices are good and your connections are clean.

Microbending-Microbends are minute deviations in fiber caused by excessive bends, pinches, and kinks. Using cable with reinforcing fibers and other special manufacturing techniques minimizes this problem.

Connector loss-Connector loss occurs when two fiber segments are misaligned. This problem is commonly caused by poor splicing. Scratches and dirt introduced during the splicing process can also cause connector loss.

Coupling loss-Similar to connector loss, coupling loss results in reduced signal power and is from poorly terminated connector couplings.

Remember to be careful and use common sense when installing fiber cable. Use clean components. Keep dirt and dust to a minimum. Don’t pull the cable excessively or bend it too sharply around any corners. That way, your fiber optic installation can serve you well for many years.

A Gigabit Interface Converter (GBIC) is a transceiver that converts digital electrical currents to optical signals and back again. GBICs support speeds of 1 Gbps or more and are typically used as an interface between a high-speed Ethernet or ATM switch and a fiber backbone. GBICs are hot-swappable, so switches don’t need to be powered down for their installation.

Although hubs and switches look very similar and are connected to the network in much the same way, there is a significant difference in the way they function.

What is a hub?
An Ethernet hub is the basic building block of a twisted-pair (10BASE-T or 100BASE-TX) Ethernet network. Hubs do little more than act as a physical connection. They link PCs and peripherals and enable them to communicate over a network. All data coming into the hub travels to all stations connected to the hub. Because a hub doesn’t use management or addressing, it simply divides the 10- or 100-Mbps bandwidth among users. If two stations are transferring high volumes of data between them, the network performance of all stations on that hub will suffer. Hubs are good choices for small- or home-office networks, particularly if bandwidth concerns are minimal.

What is a switch?
An Ethernet switch, on the other hand, provides a central connection in an Ethernet network in which each connected device has its own dedicated link with full bandwidth. Switches divide LAN data into smaller, easier-to-manage segments and send data only to the PCs it needs to reach. They allot a full 10 or 100 Mbps to each user with addressing and management features. As a result, every port on the switch represents a dedicated 10- or 100-Mbps pathway. Because users connected to a switch do not have to share bandwidth, a switch offers relief from the network congestion a shared hub can cause.

What to consider when selecting an Ethernet hub:
• Stackability. Select a stackable hub connected with a special cable so you can start with one hub and add others as you need more ports. The entire stack functions as one device.
• Manageability. Choose an SNMP-manageable hub if you have a large, managed network.

What to consider when selecting an Ethernet switch:
• Manageability. Ethernet switches intended for large managed networks feature built-in management, usually SNMP.
• OSI Layer operation. Most Ethernet switches operate at “Layer 2,” which is for the physical network addresses (MAC addresses). Layer 3 switches use network addresses, and incorporate routing functions to actively calculate the best way to send a packet to its destination. Very advanced Ethernet switches, often known as routing switches, operate on OSI Layer 4 and route network traffic according to the application.
• Modular construction. A modular switch enables you to populate a chassis with modules of different speeds and media types. Because you can easily change modules, the modular switch is an adaptable solution for large, growing networks.
• Stackability. Some Ethernet switches can be connected to form a stack of two or more switches that functions as a single network device. This enables you to start with fewer ports and add them as your network grows.

The Open Systems Interconnection (OSI) Reference Model provides a layered network design framework that establishes a standard so that devices from different vendors work together.

Layer 2 (The Data-Link Layer)
Layer 2 switches operate using physical network addresses. Physical addresses, also known as link-layer, hardware, or MAC-layer addresses, identify individual devices. Most hardware devices are permanently assigned this number during the manufacturing process.

Switches operating at Layer 2 are very fast because they’re just sorting physical addresses, but they usually aren’t very smart.

Layer 3 (The Network Layer)
Layer 3 switches use network or IP addresses that identify locations on the network. Physical addresses identify devices; network addresses identify locations. A location can be a LAN workstation, a location in a computer’s memory, or even a packet of data traveling through a network.

Network addresses are hierarchical. The more details included, the more specific the address becomes and the easier it is to find.

Switches operating at Layer 3 are smarter than Layer 2 devices and incorporate routing functions to actively calculate the best way to send a packet to its destination. However, because Layer 3 Switches take the extra time to read more details of a network address, they are sometimes much slower than Layer 2 Switches.

Layer 4 (The Transport Layer)
Layer 4 of the OSI Model coordinates communications between systems. Layer 4 identifies which application protocols (HTTP, SNTP, FTP, etc.) are included with each packet and uses this information to hand off the packet to the appropriate higher-layer software. Layer 4 switches make packet forwarding decisions based not only on the MAC address and IP address, but also on the application a packet belongs to.

Because Layer 4 devices enable you to establish priorities for network traffic based on application, you can assign a high priority to packets belonging to vital in-house applications, such as Peoplesoft®, with different forwarding rules for low-priority packets, such as generic HTTP-based Internet traffic.

Layer 4 switches also provide an effective wire-speed security shield for your network.

A media converter is a device that converts from one media type to another, for instance, from twisted pair to fiber to take advantage of fiber’s greater range. A traditional media converter is a two-port Layer 1 device that performs a simple conversion of only the physical interface. It’s transparent to data and doesn't “see” or manipulate data in any way.

An Ethernet switch can also convert one media type to another, but it also creates a separate collision domain for each switch port, so that each packet is routed only to the destination device, rather than around to multiple devices on a network segment. Because switches are “smarter” than traditional media converters, they enable additional features such as multiple ports and copper ports that autosense for speed and duplex.

Switches are beginning to replace traditional 2-port media converters, leading to some fuzziness in terminology. Small 4- or 6-port Ethernet switches are very commonly called media converters. In fact, anytime you see a “Layer 2” media converter or a media converter with more than two ports, it’s really a small Ethernet switch.