In 1965, just four years after the first planar integrated circuit was invented, Gordon Moore predicted that the number of transistors per integrated circuit would double every 18 months. In the years and decades since then, Moore's prediction has, for the most part, held true. Memory capacity, disk-drive capacity and processor speed have increased rapidly, while the costs of these components have remained constant or have fallen. This dramatic increase in processing power has not only allowed designers to create more capable computers, it has also allowed engineers to create powerful networking technologies. And powerful, high-speed networks are particularly well suited for moving the large amounts of data generated by digital video.
We're all familiar with unshielded twisted-pair (UTP) Ethernet wiring. This cabling has been used with 10 Mbit systems (10Base-T) for many years. As of 1997, 85 percent of all network connections were Ethernet, representing over 118 million connections. Given the tremendous growth in networking over the last five years, it is likely that UTP wiring is now used for over 90 percent of all network connections. Network engineers have worked hard to build on existing technologies to allow a smooth transition to new technologies while preserving as much of the infrastructure as possible. Given that, on average, over 50 percent of UTP infrastructure is still in place five years after installation. Reuse of this infrastructure is an important part of the overall plan for all the “x”Base-T technologies.
The 1000Base-T specification was ratified by the IEEE Standards Committee in June of 1999 and is formally known as 802.3abTf. It describes how gigabit Ethernet (also known as “Gig-E”) is to be carried over four balanced unshielded pairs of CAT 5 cable using V.90/56k encoding and compression.
1000Base-T's success is due in part to the large installed base of UTP cable, and it works to preserve as much of that installed base as possible. 802.3abTf expands upon earlier UTP technologies by offering the following features:
Allowing auto-negotiation between 10Base-T, 100Base-T and 1000Base-T
Continuing to use Ethernet Media Access Control (MAC) technology
Using the same clock rate as 100Base-T (125 MHz)
Providing the same robustness as 100Base-T (with a bit-error rate of 1×10-10)
1000Base-T can fit an order of magnitude more data into the same “cable” by using four pairs, each running at 250 Mbits/s. This higher data rate is achieved by leveraging the existing V.90/56k encoding algorithms (Phase Amplitude Modulation 5 and Trellis coding) rather than using 4B/5B encoding as is implemented in 100Base-T technology.
Gigabit networking is available on a number of different wire technologies — UTP is only one of the options. Gigabit networking is also implemented on optical fiber, in both single-mode and multimode, and balanced shielded copper. 1000Base-LX is the designation for single-mode fiber at lengths of up to five kilometers. 1000Base-SX describes gigabit networking over multimode fiber up to 550 meters. 1000Base-CX is the designation for gigabit Ethernet over balanced shielded copper at lengths up to 25 meters. Table 1 compares the various flavors of gigabit networks, cables and characteristics.
Backward compatibility is accomplished through an elegant combination of technologies, illustrated in Figure 1. These technologies are implementations of the Data Link Layer and the Physical Link Layer in the OSI seven-layer model.
At the top of Figure 1 is one of the cornerstones of Ethernet networking, the 802.3 Media Access Control (MAC) layer. The MAC layer takes data from an overlying application and packages it into Ethernet frames. The MAC layer is also responsible for network addressing and scheduling. Below this layer is the Gigabit Media-Independent Interface (GMII). The GMII layer provides a consistent interface to the MAC layer, regardless of whether the packets will be sent via fiber, UTP or coax.
The next layer is the Physical Coding Sub-layer. 1000Base-T uses a different encoding scheme for data on the wire compared to all of the other gigabit Ethernet technologies. For this reason, different physical coding sub-layers are introduced into the 1000Base-T stack and the stack for all other gigabit media types. 1000Base-T employs V.90/56k (Phase Amplitude Modulation 5 and Trellis) encoding, while the other technologies use 8B/10B encoding.
The Physical Medium Attachment Sub-layer consists of the electrical interfaces, and the Physical Medium Dependent Sub-layer describes the physical attachment of the wire or fiber.
When transitioning from 10Base-T or 100Base-T to gigabit Ethernet, the most important thing you can do is test your existing wiring to see that it meets the Gig-E specifications. If you have CAT 5e wiring, you can be pretty sure that the wire will meet the Gig-E specification. If you have CAT 3 wiring, you can be pretty sure that your wiring will not meet Gig-E specifications. If you have CAT 5 wiring, the only way to know is to test it. Many different cable testers are available.
When testing for Gig-E compliance, the most important measures are return loss, equal-level far-end crosstalk (ELFEXT) and near-end crosstalk (NEXT). Return loss is a measure of how well the impedance of the cable matches the impedance of the transceivers (for those of you who are ham-radio operators, think SWR). ELFEXT is a measure of unwanted signal from a near-end transmit pair that crosstalks into a neighboring pair at the far end, relative to receive signal at the far end. NEXT is a measure of unwanted signal from a near-end transmit pair that crosstalks into a neighboring pair as measured at the near end.
When migrating from lower-speed technologies to Gig-E, we really have to thank the designers of all the “X”Base-T Ethernet specifications. The advent of auto-sensing adapters, switches and other hardware has made it possible to migrate from one technology to another relatively painlessly.
It makes sense to deploy Gig-E where it is needed. The most likely places for Gig-E are areas that require very high network speed. For example, Figure 2 shows a dual Gig-E link between the switch and a high-performance file server. Also, a single Gig-E link runs from the graphics workstation to the switch to ensure that this high-bandwidth client has the maximum bandwidth available at all times. For other desktop clients who will be viewing video at a lower resolution, you can use lower-cost 100 Mbit links. Note that this bandwidth is shared among all of the desktop clients through a very low-cost hub. If at some point you need more capacity at the desktop, you can change the hub to a switch and increase the link speed between the large switch and the hub to Gig-E. Ultimately, you can change the entire network to Gig-E, but by that time, you will probably want to upgrade the large switch and the file server to 10Gig-E (but that's another story).
Brad Gilmer is president of Gilmer & Associates, executive director of the AAF Association and technical moderator of the Video Services Forum.
Send questions and comments to: firstname.lastname@example.org