At Layer 2, Gigabit Ethernet looks exactly like 10Mbps and 100Mbps Ethernet. They all apply the same 802.3 standards for framing and addressing. This similarity is convenient because it means that interconnecting Ethernet segments of these different types is simple. At Layer 1, however, the electrical signaling standards for Gigabit Ethernet are completely different.
The first set of Gigabit IEEE standards was specifically geared toward a fiber optic implementation. Naturally, the first Gigabit devices on the market all used fiber optic connectors. However, shortly thereafter, an addendum was released that included specifications for running Gigabit Ethernet over Category 5 unshielded twisted pair (UTP) cabling. Gigabit Ethernet over Category 5 cabling is called 1000BaseT. It allows for distances of up to 100 meters, similar to the 100BaseT and 10BaseT standards. This is convenient because it means that Gigabit Ethernet should, in principle, be able to operate over the same cable plant as an existing Fast Ethernet implementation.
However, there is one important caveat to this Category 5 implementation of Gigabit Ethernet, as I discuss later in this chapter. The original specifications for Category 5 cable plants did not specify signal reflection properties of connectors, which turn out to be important in Gigabit. Thus, older Category 5 cabling may not work properly with Gigabit Ethernet.
The physical layer differences between even fiber optic implementations of Fast Ethernet and Gigabit Ethernet go well beyond merely changing the clock rate. The most important issue is the use of 8B10B encoding. At its lowest level, Gigabit Ethernet uses a 10-bit byte; at these extremely high speeds, it can be difficult to accurately distinguish between bits. Thus, 10-bit patterns have been selected to represent the 8-bit octets. The specific 10-bit patterns are chosen for their transmission reliability. There is 25% of extra overhead in encoding this way, but the improvement in reliability compensates for this additional overhead.
To make implementation details easier, the Gigabit Ethernet group has defined a natural transition point in their protocol stack called the Gigabit Media Independent Interface (GMII). This sublayer is similar in concept to the Fast Ethernet MII and the standard Ethernet AUI interface. Each case specifies a point that is technically in the middle of Layer 1. Everything above this point is generic to all different implementations of the protocol. This way, only the hardware and the protocols below the dividing point need to change when a new physical layer is defined.
Most Gigabit Ethernet hardware uses either a physical fiber optic or an RJ45 connector. However, it is possible to implement a Gigabit Ethernet interface on a device using a generic GMII connector. Then the network designer could simply connect the appropriate GMII transceiver.
Now that Gigabit Ethernet is available over Category 5 cabling, putting Gigabit NICs into servers and workstations has become technically viable. The market is also already seeing price competition between NIC vendors, which drives down the costs of running Gigabit to the desktop.
I recommend using full-duplex switched connections for connecting end devices directly to Gigabit Ethernet networks; there would be little real advantage to running a shared half-duplex Gigabit network over running a switched full-duplex Fast Ethernet network. In any Ethernet environment with several devices sharing a collision domain, the effective throughput is typically 30 to 40% of the total capacity, so you can expect to get something on the order of 300-400Mbps total aggregate capacity out of a shared Gigabit hub. Each individual device on this hub would get some small fraction of this total on average. So, for a small network with 5 Gigabit devices sharing a hub, you would expect each device to have access to an average of 60-80Mbps. The peaks for each device are, of course, much higher than this, but it is reasonable to expect that devices being considered for Gigabit Accesses will be heavily used—at least in the near future.
One can already achieve a higher average utilization using simple switched Fast Ethernet. Because the cost of Fast Ethernet is still much lower than Gigabit Ethernet for both the NICs and network devices, it is not cost-effective to use Gigabit Ethernet this way.
Consequently, if an organization has end devices that are important enough and used heavily enough to warrant connecting to the network at Gigabit speeds, it makes sense to use full-duplex switched connections. As it turns out, the marketplace appears to have already made this decision, as Gigabit hubs are not made by any major vendor, while there are several vendors selling Gigabit switches.
In a unanimous decision, the IEEE 802.3 committee on 10 Gigabit Ethernet has decided not to bother implementing anything but a full-duplex version of the new protocol. So, although the standard is not yet complete as of the time of writing this book, we already know a few things about what it will look like. We know that there will be no such thing as a 10 Gigabit hub and that there will be no defined collision mechanism. This is a good thing. It seems that the market has already decided that the Gigabit standard is most useful in a switched full-duplex mode. Most organizations using Gigabit Ethernet use it as a trunk or backbone technology or attach only a small number of important servers at Gigabit speeds.
It is important to remember that this is similar to how Fast Ethernet started out. In the future, some organizations may have large numbers of Gigabit end user devices.
Adoption of the new high-speed protocol as a standard for end devices has been a little slow, mostly because of the time lag between the fiber and copper standards for delivering the medium. However, now that a version of Gigabit Ethernet that works over Category 5 cabling has been finalized and hardware vendors are releasing equipment based on the standard, there should be more use of the high-speed protocol.
The lack of a half-duplex version for 10 Gigabit Ethernet means that, when it is available, it will probably not be quickly extended to the desktop. It is not yet completely clear what sort of twisted pair copper cabling the 10 Gigabit standard will eventually use. Category 5 cable is certainly reaching its limitations with Gigabit Ethernet. However, the emerging Category 6 standard has not yet been fully embraced by the 10 Gigabit working groups, which are naturally focused on optical fiber implementations.
The bottom line is that it will be many years before you can expect to see 10 Gigabit Ethernet extended to the desktop. At the very least it will require new cable plants for most organizations, unless they happen to have optical fiber running to their desks.
I envision Gigabit and 10 Gigabit Ethernet as backbone and trunk technologies. Given the trunk aggregation rules discussed in Chapter 3, it is clear that if an organization makes extensive use of Fast Ethernet today, then it needs an inexpensive fast trunk technology. These new Gigabit and 10 Gigabit standards are ideally suited to this purpose.
One of the most positive features of Gigabit Ethernet trunks is their ability to use a common 802.3 framing throughout all levels of the network. This is important because the same VLAN tags and MAC addresses are shared throughout any Distribution Area. You don't want to have to rewrite or tunnel these pieces of information for three reasons.
First, each step introduces latency. Second, you sometimes want to put a protocol analyzer on a trunk to see what passes through it. If you can't readily distinguish the VLAN associated with a frame and if you can't easily identify the source and destination devices, it can be difficult to tell if you have a problem. Most modern protocol analyzers are able to read into a packet to help with this problem, but it can still be difficult to see what's going on, depending on the types of tunneling employed.
The third advantage to using the same 802.3 frame at each stage of a packet's journey through the network is that it ensures consistency in the treatment of its priority. As I mentioned previously, the Class of Service (CoS) field is associated with the VLAN tag. Knowing this allows the network to have a consistent logical identifier for the prioritization scheme to use at each hop up until the packet hits a router. At the router, of course, a higher layer identifier (such as the IP TOS or DSCP field) has to carry the information, since the packet will lose its Layer 2 information as it crosses through the router.
I consider Gigabit and 10 Gigabit Ethernet naturally suited to trunk links in large-scale LANs. Interestingly, much of the current discussion regarding these standards involves their use in larger Metropolitan Area Network (MANs) and Wide Area Networks (WANs) as well. As it is currently common to see MAN and WAN networks implemented using ATM and delivered to the customer premises as an Ethernet or Fast Ethernet port, it does seem natural to extend this delivery to Gigabit speeds as well. Certainly this extension would give efficient near-capacity access to the current highest-speed ATM link technologies. It might turn out to be a good low-cost delivery mechanism for these links. However, any more detailed discussion of WAN technologies or speculation on yet unwritten standards is beyond the scope of this book.