The amount of data traffic coarsing around networks the world over continues to increase, so much so that it’s become a struggle for many systems to keep up with the pace. Ken Mann reports on how Gigabit Ethernet can give your network the green light it so badly needs

The arrival of fast PCs and high speed LAN applications, combined with the growing interest in multimedia, have provided the means and incentive to dramatically increase network traffic. With these conditions it is inevitable that developers will create applications to capitalise on bandwidth availability, just as they have capitalised on higher CPU performance.

A factor accelerating this trend is the changing style of business management. Hierarchies are becoming flatter and staff at lower levels are being empowered to make more decisions. To do this they must have fast access to more of the corporate information resource. As a result, many companies are now implementing bandwidth hungry intranet systems to make information systems easier for staff to use, and speed the flow of information through the organisation. The savings in travel costs and time offered by business quality video, audio, and data conferencing are also becoming harder to ignore. When these communication intensive applications finally arrive, their demands on network capacity will dwarf that of most current
business software.

Yet another factor driving up bandwidth demand is the introduction of network PCs or "thin clients", aiming to bring down costs and ease network administration. By storing application software, user environment information and files on high performance servers, the network PC increases LAN dependency and makes far greater demands on network capacity.

The need for speed
Comments James Donovan of Lucent Technologies, "Even if such innovations generate only 10% of the PC's network traffic potential, at current rates of processor speed increase, gigabit speeds to the desk will be a requirement within six years. Over this time, traffic patterns will also change. The use of network PCs, for instance, can be expected to alter the traffic balance from 80% local workgroup to 80% remote, resulting in disproportionate growth in loads on backbone systems."

Among recent cabling developments is the GigaSPEED structured cabling solution originating from Bell Laboratories. This is the most important development in structured cabling at Bell Labs since its SYSTIMAX group introduced the Category 5 SCS in 1990. Similarly, MiTech has launched System 6, a fully compliant Category 6 cabling solution.

Although Cat 6 is not yet a ratified standard, Siemon warrant that System 6 performance significantly beats existing specifications and will exceed the standard requirements. System 6 becomes MiTech's standard cabling solution, with System 5E used only in specific situations - for example, where customers have temporary occupancy of premises. Priced to compete with Cat 5 systems, System 6 offers two-and-a-half times the bandwidth of the former for only 15%-20% additional investment.

"A few years ago nobody would have imagined that we would be looking at technology of the scale of Gigabit Ethernet. Yet this is now the accepted standard and we are currently anticipating further advances. The cabling infrastructure needs to continue to grow with these developments and we are keen to stay ahead of the pack," comments Eric Thickett, MD of MiTech.

"70% of computer network downtime is due to poor cabling," said Nick Norris, sales and marketing manager at International structured cabling specialists KRONE, which has recently been acquired by GenTek. Quoting from published research (BSRIA Network Survey), Norris went on to say "few companies give a second's thought to the cabling that connects their computers together. They spend weeks and months agonising over the type of PCs, servers, routers, hubs and data-backup systems and then get some local electrical outfit to install the cabling." They don't realise how sensitive computer cabling is - because it looks just like phone cable -but it's a bit like asking a butcher to perform open-heart surgery!"

The result is a computer network held together with the electrical equivalent of string and sealing wax. It'll work OK on day one, but as the cabling system beds-in and starts to age a little, then one or two sockets are moved around a bit you'll find that more and more computer problems start to occur. But you won't know why and because it's gradual you may not even realise it's happening. But you will find more computer faults - you'll maybe need to employ another technical support person or two as time goes on."

Launching KRONE's new PremisNET structured cabling package Norris said "We wanted to take all of the risk out of deciding which cabling system to choose - to make life simpler for end-users and specifiers. So we've come up with a really straightforward solution.” PremisNET is a complete infrastructure solution, which bundles together fibre, fibre connectivity, copper cable, copper connectivity and the KRONE warranties. Naturally, when a PremisNET solution is installed by a KRONE Associate Installer, the complete end-to-end performance warranty will apply, which is now extended to a full twenty years worldwide. There will initially be four performance levels in the PremisNET portfolio:
• Bronze is the entry level Cat 5 100MHz solution typically supporting
10Base T and 100Base T.#
• Silver is the Gigabit solution and takes on board the new 1998 Cat 5
specification, which defines extra performance parameters. Silver is
suitable for Gigabit Ethernet and 155/622 ATM.
• Gold is a "Cat 6 channel" solution, which uses a combination of Cat 6
cables and KRONE's high headroom Gigabit capable connectors to extend the application range to include 2.5 Gigabit Ethernet.
• Platinum is KRONE's fully fledged Cat 6 solution, using the new KRONE Cat 6 connector and Cat 6 cables.

Differentiated Services
Only three years ago, most TCP/IP applications were well served with the
best-effort quality of service (QoS) delivered by TCP/IP. With the tight
design constraints of latency (delay) and jitter for voice over IP, coupled
with large file transfers from File Transfer Protocol (FTP) or Hypertext
Transfer Protocol (HTTP) applications, and with the typical domain name service (DNS) lookups and Simple Mail Transfer Protocol (SMTP) mail traffic, networks today must support a wide range of applications. Today's HTTP and HTML-based Web applications account for about 80% of the traffic on the Internet. By nature HTTP is a bursty protocol - but voice over IP is not.
Integrated Services bases much of its promise on RSVP. As one application
example, RSVP is currently the preferred voice-over-IP QoS signaling
mechanism for enterprise private WAN backbones. What makes RSVP appealing is that it delivers guaranteed bandwidth. For applications such as voice over IP or IP-based video streaming or conferencing, which are initiated by either an end station or the network, RSVP provides call admission by signaling end to end from the source to the destination. To guarantee the necessary resources, this signaled approach sets up a path from end to end by negotiating the requested parameters held within the Tspec and flowspec on each network router and Layer 3 switch along the path. RSVP offers a unique benefit by ensuring that the network does not send any excess voice-over-IP traffic down a link or trunk and that bandwidth is available so that another call can be accepted along the identified path.

Differentiated Services (Diff-Serv) defines a framework for IP aggregation that delivers end-to-end network performance. Several RFCs already lay out the basic framework for Diff-Serv. These RFCs expand on the delivery of end-to-end QoS by introducing two new concepts:
• Expand the number of available IP-layer classes to 64.
• Define the association of the marking of the packets to a predefined
(minimum) behaviour within each network device along the forwarding path.

The first of these goals is relatively simple to achieve. Today's precedence is based on the eight-bit IP version 4 (IPv4) type-of-service (ToS) field. Within this field lie three bits slated for IP Precedence. These bits provide a user-definable range of services from a low of zero to a high of five, with six and seven reserved or preferred for use in management and routing updates.

The Diff-Serv Code Point
Diff-Serv redefines this ToS field while providing backward compatible with IP Precedence. The new ToS field definition supersedes the IPv4 (RFC 1349) and IPv6 headers. Six bits of the Diff-Serv field are used as a code point (known as the Diff-Serv Code Point, or DSCP) by each Diff-Serv-aware switch or multiprotocol router along the forwarding path to invoke the defined behaviour. Two bits are currently unused and are labeled as such (CU) by Diff-Serv, but are claimed in RFC 2481, "A Proposal to Add Explicit Congestion Notification (ECN) to IP" (January 1999). Six bits of the Diff-Serv ToS field are used as the Diff-Serv Code Point, or DSCP, while the remaining two bits are currently unused. The code point space is divided into three pools for assignment and management. Within these three pools, RFC 2474 ("Definition of Differentiated Services Field") assigns eight recommended code points from pool one to provide backward compatibility with IP Precedence. This capability provides a powerful migration path today for end-to-end deployment of IP Precedence-based QoS.

With the Diff-Serv Code Point, it's important to mention an additional concept, the per-hop behaviour (PHB). PHB is the observed behaviour of packets at each network element, such as a router or a Layer 3 switch, when multiple requests vie for element resources, such as buffer and bandwidth. The IETF standard defines two PHBs:
• Expedited Forwarding (EF)
• Assured Forwarding (AF)
EF translates into the required behaviour for voice-over-IP service: low loss, low latency, low jitter, and available bandwidth. The traffic conditioners configured today to deliver this level of end-to-end QoS parameters include:
• Priority Queuing
• Class-Based Weighted Fair Queuing (CB WFQ)
• Weighted Round Robin (WRR)
• Link Fragmentation and Interleaving (LFI) - Compressed Real-Time Protocol (CRTP) - Weighted Random Early Detection (WRED)

Latest developments
Gigabit Ethernet network solutions have been developed in response to these changing demands on the network. For example, Foundry Networks, the first company to ship a fibre Gigabit Ethernet Layer 2/3/4 switch, have introduced a suite of copper Gigabit Ethernet products, including a high density 1000Base-T switch, as well as 1000Base-T modules for existing Foundry products.

The switches are part of the FastIron II family of edge switches and provide customers with a range of high-density 10/100/1000 Mbps solutions for wiring closets, server farms and workgroups. Enterprises can now deploy Foundry's copper Gigabit Ethernet technology for approximately $650 per port, which is significantly lower than the average per port price of fibre Gigabit Ethernet.
Desktop switch players such as D-Link have also added Gigabit Ethernet to their ranges. The D-Link DES-5008 is a single port Gigabit Ethernet Module for the DES-5024. Its 1000Base SX Gigabit port with a SC connector allows seamless connection to another Gigabit switch, or Gigabit NIC to increase inter-device bandwidth to 1000 Mbps.

The D-Link DES-5024 is a high performance 24-port 10/100Mbps Ethernet/Fast Ethernet NWay auto-negotiation modular Switch with SNMP/RMON management, and optional Gigabit up-link connections and ISDN Router.

The FastIron II from Foundry Networks. Foundry have just launched a range of copper Gigabit Ethernet products

Above : Krone PremisNet Highband patch
-by-exception is an alternative to RJ45 patching systems for CAT 5R and CAT 6

Above : The DES-1024 10/100 Fast Ethernet Switch from D-Link