Over the past several years our industry has been distracted, yes distracted by the dotcom bust, telecom crash, global recession, bankruptcies and executive scandals. During this period, the networking industry has continued to innovate and produce new technologies and products. While we had or heads down working hard reacting to external pressures and threats, the networking industry changed. Over the past three years the networking vendors have financed billions in R&D investment. What we got for that investment is network security products, IP Telephony, wirelesses LANs, 10G Ethernet, storage area networking and broadband. All these new technologies are now reshaping enterprise networks in ways that were unanticipated. For example, a new two-tier LAN structure made up of smart workgroups and a distributed core backbone is emerging.

As a result of this investment, all the assumptions and enterprise network design rules from the 90s have changed. How we design voice networking is totally different now as are LANs, wide area networks and network services. An excellent example of this change is the WLAN design tool Trapeze Networks has developed to assist network managers in access point placement, spectrum coverage, channel assignment, etc.

At the same time many IT executives struggle to get up to speed on new design techniques, high on enterprise’s list of funded projects are hardening the infrastructure through network security, identifying vulnerabilities and developing disaster planning
scenarios to assure business continuity. Clearly network security and business continuity are important projects for these times. But there is one problem with these projects; they are inflationary to IT budgets meaning that they don’t provide one
additional dollar toward revenue/employee productivity. In short, security and business continuity projects allow us to sleep better at night but they don’t do anything to increase revenue or earnings. They too present new design approaches and challenges to enterprise networks.

Network design is changing is because corporations needs have changed. We are entering an era where enterprise networks allow companies to be more adaptive and responsive to their customers and suppliers plus transparent to shareholders and
investors. In this Lippis Report we’ll explore the new two-tier DIN LAN structure. Companies that are building product for a two-tier model are HP’s ProCurve Group www.hp.com, Avaya www.avaya.com, 3Com www.3com.com, Enterasys www.enterasys.com, Foundry www.foundry.com and Extreme www.extreme.com. While Cisco www.cisco.com champions a three tier LAN structure it clearly has the products and motivation to shift to a simpler two-tier model. We can add Nortel www.nortel.com to that list of DIN supports as well after a discussion I had with Greg Merritt, VP Enterprise Networking Marketing at Nortel.

Three-Tier LAN Structure

The traditional three-tier LAN structure includes low cost workgroup switches to distribute access to desktops. This distribution tier connects to an aggregation tier made up of modular LAN switches in a wiring closet. The aggregation tier then connects to the third core backbone tier. The third tier is made up of a large chassis switch usually located in a data center or facilities room. The third tier is sometimes referred to as a collapsed backbone with all traffic flows transiting the core switch backplane. In essence the switch backplane provides the key bandwidth for inter-LAN segment traffic flows. The collapsed backbone connects into a router for wide area connectivity. Over the years the collapsed backbone switch has been the platform where most network services and features emanated from, burdening the core switch with endless software releases and fatten feature sets of questionable value. In essence this feature load on core switches invariable increases its cost and complexity. Backplane speeds in collapsed backbone switches have been able to keep up with traffic demands, but all traffic is forced to flow through the collapsed backbone, which may not be the best path for optimal user performance.

Today’s core switched networks are dominated by legacy chassis based designs, which offer high cost of acquisition and complexity. During the mid to late 90s legacy chassis designs offered a degree of freedom to IT managers as LAN standards were in consent flux. While the networking industry was working its way through various LAN standards such as token ring, versus FDDI, versus Ethernet versus ATM plus shared versus switched networks, IT managers needed a way to migrate and transition between these standards. Legacy chassis offered a network design that allowed network managers to experiment and change directions with the change of a module. IT managers were willing to pay the high price of a legacy chassis in exchange for the flexibility to change LAN standards and migrate between them. Today’s LAN market is different. Ethernet is the standard and the need for LAN type experimentation is a thing of the past. There is no need to pay the high price of legacy chassis solutions just to hedge on a LAN standard. Ethernet won that war. During this process IT managers were trained to pay high prices for core switches. Well its time to get re-educated.

Two-Tier LAN Structure

The two-tier structure includes smart workgroups and a distributed core made up of modular/chassis based products. First, the role of the core backbone network is changing to one that provides a highly reliable transport service. Why do network services have to be centralized in core backbone switches? In effect, as workgroup networks become smarter with the increased intelligence, core networks become simpler and less expensive. Intelligence is moving to workgroup switches in part to stem off the onslaught of firms like Dell www.dell.com, Linksys www.linksys.com, Intel www.intel.com, et al, who offer nearly feature free switches to the small enterprise market. As the traditional equipment suppliers add network services to workgroup switches, a barrier of entry is added, capping the commodity players from participating in large enterprise networks. Another reason is network security features need to be in smart workgroup switches to close access vulnerabilities of both wired and wireless
connections.

A new simpler core design with distributed switching is emerging. This core design is not based upon the exclusive use of chassis based switching placed into a collapsed backbone structure, but on distributed switching delivering lower cost of acquisition and increased network design flexibility. Figures 1 and 2 illustrates two and three tier structures respectively. In figure 2, the distribution layer has been omitted. Network services such as VLAN tagging, load balancing, WLAN security, roaming, QoS, port security, Power over Ethernet etc, identified as circles of color in figure 1 & 2 are delivered through a centralized network
management system in distributed core networks. Notice how services shift from the collapsed backbone switch toward smart workgroups with core networks providing key transport, reliability and service distribution.

Two or Three-Tier LAN Structure?

With decreasing price points of switching across the packaging spectrum core network devices can now be cost effectively distributed. This avoids network traffic bottlenecks and increases reliability by eliminating centralized single points of failure inherent in collapsed backbone networks. From a cost point of view, core networks designed with distributed switched networking products are less expensive to acquire since IT managers purchase what they need when they need it. Also, as requirements increase, these distributed core networks scale up more cost effectively thanks to their improved price/performance metrics over large chassis. For high port density applications, chassis based switching solutions are most effective and are an important part of the new distributed core design. New Ethernet switch chassis offer a wide range of Ethernet standards such as 10/100/1000 Mbs, called triple speed, and high port density. Today’s chassis products fit in and work within a distributed core network that
distributes not only packet forwarding but service delivery too. Most importantly, there is an alignment between traffic flows and physical paths since tier two designs eliminate the requirement that all traffic pass through the collapsed backbone. This increases performance by providing a more direct path between source and destination. The distributed core can be made up of fixed configuration or chassis based switches. A key feature set of distributed core backbones are that link aggregation is common between devices allowing dynamic allocation of high-speed links to heavy destination spots. To minimize operational spend routing and management of all devices within the distributed core backbone is represented as a single image. Companies’ operations and business processes are changing. This change is occurring most acutely at the user level. Employees need to be empowered by their IT infrastructure to adapt to a new mode of operations. The two-tier DIN LAN structure with network services closest to users starts to deliver that agility.

Posted 30 June 2003 at 7:12 pm under Lippis Report, Enterprise Mobility. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.