Standards and Gigabit Transmission
|In general the networking industry is developing gigabit transmission based on standards. This newsletter article summarizes the most important standards.|
In general, the industry as a whole is developing gigabit transmission based on standards. The Institute for Electrical and Electronic Engineering (IEEE), which should be familiar to you as the main source of LAN standards, is extending its work into gigabit campus and metropolitan systems. Longer-range systems are part of optical work in the International Telecommunications Union (ITU), Metro Ethernet Forum (MEF), and, to some extent, the temporary sub-IP area of the Internet Engineering Task Force (IETF).
Since gigabit networking often intersects optical networking, it will eventually come under the evolving Generalized MPLS (GMPLS) standards work of the IETF. GMPLS extends the MPLS setup with which you are familiar to set up paths for other than packet-switched communications, such as time-division multiplexing and optical wavelength multiplexing.
Gigabit systems exploit sublayering in the IEEE 802 architecture. This sublayer may not be immediately familiar, or you may know it by earlier names. While the most common Ethernet implementation today is star-wired 10/100BaseT, the original 10Base5 implementation separated the computer equipment from the actual medium interface. The medium interface was on a medium-dependent box called a transceiver, which was connected to the computer by an attachment unit interface (AUI) cable. You will see the 15-pin AUI interface on older Cisco equipment.
10Base2 "thinwire" implementations still separated the Physical Layer Signaling (PLS) part of the physical layer from the Medium Independent Interface (MII), but the connection between the two would usually be internal to the computer. Transceivers are available; however, to connect AUI interfaces to 10BaseT cable.
PLS was at the top of the OSI physical layer. In the newer and faster Ethernets, its function is called the reconciliation layer, again at the top of the physical layer, and translates layer 2 octets into layer 1 bits. It may introduce overhead bits that have no direct correspondence to data bits.
Figure 1. IEEE Stacks
Where an AUI interface was previously used to get to a medium-specific interface, Gigabit Ethernet uses the Gigabit Ethernet Interface Converters (GBIC) as the transceiver. The GBIC plugs directly into the chassis; there is no GE equivalent to the AUI cable.
The "far side" of the GBIC, with reference to the Reconciliation Sublayer, is the physical coding sublayer (PCS). In the original GE standards, it converts 8 data bits to 10 medium bits (i.e., 8B/10B encoding). You may remember that FDDI uses 4B/5B encoding, with the actual medium rate 120 Mbps and five medium bits for every four-bit data symbol. These overhead bits guarantee enough signal pulses (or more specifically, transitions) for the receiver to recover clock.
A key driver in getting these technologies to market is that they used physical medium dependent chipsets already in place for other standards. The first 1 Gbps Ethernet used ANSI X3T11 Fibre Channel. 10 Gbps Ethernet uses the same chipset as SONET OC-192.
Standards for Ethernet-family protocols generally come from Project 802 of the Institute for Electrical and Electronic Engineers (IEEE).
Table 1. IEEE Standards and Gigabit Networking
|802.1 Architecture||802.1d||Basic spanning tree|
|802.1p||Quality of Service|
|802.1w||Fast Spanning Tree|
|802.1ad||[Service] Provider Bridges|
|802.2 Layer 2||802.2||Logical Link Control|
|802.3 CSMA/CD Physical||802.3||Basic 10 Mbps|
|802.3z||Gigabit Ethernet over Copper|
|802.3ab||Gigabit Ethernet over Fiber|
|802.3ad||Inverse multiplexing (EtherChannel equivalent)|
|802.17 Resilient Packet Ring||See High Availability Study Guide||More bandwidth-efficient alternative to traditional SONET/SDH protection switching, and even the somewhat more efficient techniques of Next Generation SONET|
Even faster speeds are coming. 40 Gbps OC-768 is beginning to get out of the laboratory, and there is now an IEEE working group for 100 Gbps Ethernet.
A slight exaggeration
10 Gbps Ethernet doesn't really run at 10.000 Gbps, but at a respectable 9-plus Gigabit rate. For many applications, OC-192 and 10 Gbps Ethernet run over the same chipsets, as will OC-768 and 40 Gbps Ethernet -- which is really 39-plus Gbps.
In the practical Cisco world, however, it's worth paying attention to "plain, ordinary" Gigabit Ethernet, because it differs from slower Ethernets by physically plugging into equipment, and having some protocol and timing variations. Since the higher-speed versions on copper and fiber are full duplex, the CSMA/CD aspect of basic 802.3 does not come into play. You may, however, see 802.3x flow control.
The new speeds are all thoroughly standards based, with the basic references being:
802.3z and 802.3ab - 1 Gbps Ethernet. IEEE 802.3z defines GE over copper, while 802.3ab defines GE over fiber.
802.3ae - 10 Gbps Ethernet
There are standards not for the medium, but for how it might be used, especially in large campus or metropolitan area applications:
Additional standards, which you are likely to encounter only if you are studying optical technology for CCIE SP or CCNP specialist exams, include a market trend toward Metro Ethernet as the dominant means of customer interconnection, and, in coordination with other technologies, becoming the basic SP technology. SP internal use is evolutionary, running Ethernet framing over PPP-over-SONET (POS), SONET, SDH and the new Optical Telecommunications Network (OTN)/Advanced Switched Optical Network. The Metro Ethernet Forum (MEF) has been the focus of practical industry work in operational use.
In the International Telecommunications, there is an active "Ethernet over Transport Architecture (G.eota)" paralleling the many other groups working on making Ethernet framing independent of the underlying transport.
I said "transport", not "transport"
People with data backgrounds can easily be confused by the terminology of these groups. When they speak of "transport", they are not referring to layer 4 of the OSI Reference Model. In OSI terms, they are referring to functionality not above Layer 2, and frequently in the transmission system below Layer 1.
It is a nuance often ignored in Cisco classes, but Layer 1 of the OSI Reference Model really refers to the connection to the medium, not the medium itself. You might use any of several interfaces to connect to a DS-1 signal, but those interfaces are unaware of the multiplexing to DS-3, ATM, POS, etc., inside the service provider
Table 2. ITU Ethernet standards work
|G.eota||Ethernet over transport architecture|
|G.enni||Ethernet over transport network interface|
|G.esm||Ethernet over transport Ethernet service multiplexing|
|G.ethna||Ethernet-layer network architecture|
|G.ethsrv||Ethernet over transport service characteristics|
|G.euni||Ethernet over transport user interface|
|G.smc||Service management channel private line|
Before going into Gigabit and Ten Gigabit Ethernet, you may find it helpful to review some optical technologies used by these systems. Even before that, let's quickly review all the media standards used, both copper and fiber.
For GE, think 1000BaseQQQ, where QQQ is the medium type. Remember 1000BaseT is the only form that can autonegotiate speed, letting today's Fast Ethernet cable serve tomorrow's GE.
Unfortunately, of the QQQ values, only T and CX have consistent meanings:
T is Category 5 or better copper. Categories 5 and 6 are unshielded.
CX is Category 7 shielded copper.
Note that there are more than pure electrical characteristic differences between Cat 5 and higher levels: at Cat5e and higher, you must have all eight wires terminated. Cat 5 can have only four wires, although eight is perfectly permissible.
Things get more complex when you move to optical media. Since this tutorial is focused on Ethernet in enterprise applications, we need not get into the quite complex area of optical fiber types. However, there are several basic things you do need to know, such as the two basic kinds of optical fiber.
There are two basic kinds of optical fiber: multimode (MMF) and single-mode (SMF). Each has subtypes beyond the scope of this paper, such as modal bandwidth, or special optical manufacturing methods such as stepped or graded index. For campus application, you should be concerned with multimode versus single mode, with 50-micron multimode having greater capability.
In and of themselves, you cannot judge the range just by the fiber type, but also need to consider the light source and its wavelength. Basically, light-emitting diode (LED) light sources are cheap but short range and bandwidth limited, while lasers can be long-range and fast, but are more expensive.
You will see distinctions made between "wavelength" of a fiber and "reach" or "haul" of a fiber. Wavelength isn't really a characteristic of the fiber itself, but of the light source. 1000BaseSX is short wavelength and 1000BaseLX is long wavelength GE.
Long laser wavelengths can drive signals farther, but they are more expensive, and they work best over more costly single-mode fiber. Neither short or long wavelength, nor single-mode or multimode, is "best". The choice is a tradeoff between cost and functionality. If you have to go only a few meters, it's silly to use a more expensive technology when a cheaper one will do the job perfectly well.
Table 3. Reach
|Type||Typical Range||Medium Types|
|Very Short Reach (VSR)||10-15 meters||Copper or multimode fiber. Parallel conductors for 10GE. Copper is Cat 5 or better, Cat 7 for 10GE.|
|Short Reach (SX)||In the data center and nearby, Hundreds of meters to 1-2 kilometers||Multimode fiber|
|Intermediate Reach (IR)||< 10 KM||Multimode or single mode|
|Extended Reach (XR)||10-50 KM||Single mode|
|Long Reach (LR)||< 50 KM||Single mode|
|Ultra Long Reach||Tens of thousands of kilometers, as in transoceanic cables.||Advanced single mode|
While basic optical connections run on a single wavelength over "dark fiber", increased capacity requirements, especially for service providers, lead to the use of Wavelength Division Multiplexing (WDM) on appropriate cables. WDM allows you to use several wavelengths simultaneously. You can think of each wavelength as a "virtual fiber".
Wide WDM (WWDM) is a Cisco technique that lets you run 10 GB over less exotic fiber than normally required. Think of it as "Virtual Gigabit EtherChannel", in which the 10 Gbps signal is made up of four 2.5 Gbps signals, each on their own wavelength.
As opposed to the inverse multiplexing of WWDM, Coarse (CWDM) and Dense Wavelength Division Multiplexing (DWDM) run multiple independent streams. They are principally for long-haul applications, but occasionally useful for expanding the capability of an existing campus fiber plant, allowing you to put multiple gigabit and 10-gigabit signals, without mutual interference, on a shared medium. CWDM is cheaper and simpler, but does not offer the capacity of DWDM. CWDM may be more than adequate for campuses or for linking campuses through leased fiber.
Table 4. WDM Components
|MSSP||Multiservice Switching Platform||Supports all topologies, optical cross-connect, and multiplexing functions|
|MSPP||Multiservice Provisioning Platform||Edge device that participates in different multiplexing techniques and protection topologies|
|MSTP||Multiservice Transport Platform||Edge multiplexer. Supports different multiplexing modes but limited protection technologies.|
Several switches and routers have Gigabit Ethernet Interface Converters (GBIC) that assign the GE signal to a particular wavelength of light.
Table 5. Cisco Standalone WDM Products
|Device Type||SONET||SDH||SONET SDH|
|DWDM MSTP||15454 MSTP|
|CWDM MSTP||Often implicit in CWDM chassis on switch|
Did you say "Color"?
You can imagine the actual wavelengths, or lambdas after the scientific symbol for wavelength (?), as much like the bands of color into which a prism splits white light. That analogy will go only so far, as the actual wavelengths used are in the infrared range and invisible to human vision.
Being invisible, however, doesn't mean that they can't affect the human eye. Ignore warnings at your peril about not looking into fibers, keeping safety shutters in proper operating condition, and putting covers over active interfaces or fibers not plugged in.
It is urban legend that some optical interfaces have the warning "do not look into connector with remaining eye", but the story makes an important safety point. Even lower-intensity optical interfaces can cause eye damage. Remember that the wavelengths used are invisible to the human visual apparatus, but very real dangers.
You will find the term "lambda" used two ways: as a general term for a wavelength, but also as the term for a specific beam of light as it works its way through the optical network.