The Smart Choice for Token Ring Networks
With the majority of Token Ring users having designed a switched LAN environment that incorporates a mixture of Ethernet and Token Ring transport, network managers are focusing their efforts on how to move switching to the desktop, while ensuring the coexistence and interoperability of Token Ring and Ethernet workstations over the corporate backbone
Over the past year, a number of vendors have incorporated significant advances in the functionality of their stackable and chassis-based second generation Token Ring switch product lines. Network managers can now build Token Ring networks that provide all the benefits of switched versus shared media at an affordable price. The best of these powerful new switch offerings have provided ASIC-based, media speed switching support for all the bridging modes, Dedicated Token Ring (DTR) attach capability for all switch ports, high port density, media speed filtering and Remote Monitoring (RMON) management. These switches can be used for backbone switching, workgroup microsegmentation and dedicated Token Ring switching to the desktop.
However, the network landscape is changing rapidly, and despite the welcome advances that have been made in Token Ring switching deployment, the majority of Token Ring users have designed a switched LAN environment that incorporates a mixture of Ethernet and Token Ring transport. Top among the reasons for constructing mixed LAN infrastructures is the rising number of corporate mergers and acquisitions, today’s attractive pricing of Ethernet network interface cards (NICs) and switch ports, and a need for flexibility in the network to facilitate rapid deployment of new applications. Network managers are squarely focused on how to move switching to the desktop, while ensuring the coexistence and interoperability of Token Ring and Ethernet workstations over the corporate backbone.
Mixed LAN Media Coexistence, Interoperability and Migration
The importance of mixed LAN media switched transport is highlighted by a recent study of Token Ring user in the United States and Europe. The study showed that only 22 percent of Token Ring users planned to maintain a pure Token Ring switched infrastructure, while 26 percent are aggressively migrating to Ethernet, and 52 percent plan to integrate Token Ring and Ethernet transport over a single campus network. It seems that no matter which of the three migration scenarios is implemented, network managers have chosen to combine the functionality of second-generation Token Ring switches with the following key architectural design elements:
*Switching to the desktop
*Multi-Virtual LAN (VLAN) transport capability
*Media independent backbone switching for Token Ring and Ethernet traffic
Switching to the Desktop
There is a popular misconception in the industry today that the only reason to push switching to the desktop is to provide greater bandwidth to the end-station. In actuality, there are two primary reasons to attach end-station client and server devices directly to the switched network. The first is to provide dedicated bandwidth for delay sensitive, multimedia applications. Whether the network is using Token Ring or Ethernet media access protocols, the bandwidth to an end-station directly attached to a switch port should be dedicated. End-stations having dedicated bandwidth give network managers complete application deployment flexibility. No matter what bandwidth is needed by the application, it does not have to contend with other devices on a shared media to prioritize its traffic above that of other stations. Isn’t the real purpose of the network infrastructure to give end users the ability to utilize the latest application software to do their jobs effectively? The needs of the business should dictate the speed of network changes and new application deployment, not the limitations of the LAN topology.
The second reason to extend switching to the desktop is to diminish the scope of network outages caused by devices connected via hubs in a shared media environment. For example, a beaconing condition on a shared media Token Ring segment or a jabbering station on a shared media Ethernet segment will prevent all stations attached to that segment from communicating. However, the network impact of the beacon or jabbering station in a dedicated switched environment is limited to only the directly attached station. Similarly, with high numbers of soft errors, all stations attached to a shared media segment experience degraded performance. In a dedicated switch port environment, only the one end-station would be impacted. Furthermore, switches that support RMON and SNMP management protocols can forward the problem "fault domain" to the SNMP management station for fault identification, determination and correction.
Replacing Hubs with Switches in the Wiring Closet
In order to replace hubs in the wiring closet with switching to the desktop, the price of a dedicated switch port must be comparable to the price per port of the chassis-based, shared media concentrator modules. Likewise, the switch ports must perform ring-speed auto-sensing and ring number learning - functions found in intelligent hubs today.
In addition, full conformance to the IEEE 802.5r standard is necessary to support direct attachment of an end-station as well as the capability to negotiate half- or full-duplex data transmission. Finally, fiber Token Ring switch ports need to support the IEEE 802.5j standard signaling for ring-in/ring-out (RI/RO) and for fiber lobe extension, as well as IBM’s proprietary fiber signaling used on the 8230 Controlled Access Unit. This allows multiple switches in the wiring closet to be connected into a single logical ring segment to maintain the network’s existing ring numbering plan.
Multi-VLAN Backbone Transport
The ability to configure and extend Virtual LANs (VLANs) across the campus LAN infrastructure helps administrators who can no longer scale a single flat LAN broadcast domain to connect all clients and server devices in the network. The main reason to define multiple VLANs in the network is to limit the number of devices in a single broadcast domain. The larger the number of clients and servers in a LAN broadcast domain, the larger the number of broadcast frames that are generated and propagated through the network. In medium to large Token Ring LANs, the use of fast LAN switches instead of traditional bridges can potentially result in a broadcast meltdown as the number of broadcast messages forwarded by the switches overwhelms the end user stations. Using VLANs to control the scope of the broadcasts in the network in tandem with inter-VLAN bridging/routing intelligence to pass traffic between the VLAN domains is a proven method for scaling even the largest campus LANs.
A good way to use multiple VLANs is to base their size on the size of the network’s primary routable protocol (i.e., IP) subnet mask. For example, if a subnet mask is 255.255.255.0, define multiple VLANs so that each has about 250 devices per VLAN. This approach allows for maximum utilization of the IP addresses and results in broadcast domains of a manageable size.
Movement of computers from one subnet to another can be managed via Dynamic Host Configuration Protocol (DHCP). Almost all of the major operating systems support DHCP today. Cisco provides a DHCP and Domain Name System (DNS) server product that automatically updates DNS when computers move to a different subnet. Moves for SNA, NetBIOS and IPX clients are generally not a problem because they either automatically reconfigure when moved or keep the same network address.
Cell and Frame-Based VLAN Transport Options
For enterprises who have a mix of Ethernet and Token Ring or for enterprises who have a heavy concentration of routable protocols (IP, IPX, AppleTalk, Decnet, etc.), the ability to multiplex VLANs over high-speed backbone connections is very desirable. For Token Ring, two methods of high speed, multi-VLAN transport are available today – ATM and InterSwitch Link (ISL).
The ATM Forum defined LAN emulation (LANE) to provide support for multiple VLANs over ATM. The ATM LANE specification supports transport of both Ethernet and Token Ring traffic, as well as support for source route bridging (SRB) for Token Ring LANE. In an SRB environment, the emulated Token Ring LAN (ELAN) appears as a single ring segment in a VLAN with its own unique ring number. An ATM port on a Token Ring switch can be connected to multiple ELANs. Each connection acts like a LAN port attached to a logical ring (ELAN).
Therefore, the ATM VLAN model for source routing requires two bridge hops to cross the ATM cloud. The ELAN appears as a ring with source-routing bridges at the edge of the ATM cloud to connect to the legacy Token Ring segments. Servers and other devices can also be attached to ATM and be a part of the ATM ELAN.
As shown in the figure above, InterSwitch Link (ISL) is a second method of providing multiple VLANs over a high-speed connection. ISL is a tagging approach for support of multiple VLAN transport over 100-Mbps and 1-Gbps Fast Ethernet media. ISL provides a frame-based solution that consists of a series of point-to-point links used to traverse the campus network. The ISL network appears as a single, distributed source-routing bridge. Therefore, there is a single hop from Token Ring port to Token Ring port using ISL technology, regardless of the number of ISL switches in the middle of the network. SRB support means that ISL links can support parallel active bridge paths between segments. In addition to SRB, the Source Route Transparent and Source Route Switching bridge modes are also supported using ISL. A spanning tree instance is run for each VLAN carried on the ISL trunk link to prevent loops in the network.
Why Frame-Based Backbones?
ISL provides a less expensive and less-complex alternative to ATM LANE for a campus backbone solution. Because it is a series of point-to-point switch links, ISL does not require an ISL switch as do ATM LANE configurations. As a result, the overall cost per seat in the network can be reduced. With ISL there are no overhead LANE services (LAN Emulation Configuration Server, the Broadcast and Unknown Server, and the LAN Emulation Server) required to configure the backbone transport. This technology, long used in Ethernet networks, has been extended to support Token Ring VLANs as well. For Token Ring VLANs, the full duplex, 100-Mbps ISL links can support full-length Token Ring frames, including SRB information. Translational bridging from Token Ring to Ethernet is not required because the entire Token Ring frame remains intact when transported across the 100 Mbps link. ISL can be used as a transport between switches and between switches and ISL-attached servers. This provides an excellent vehicle for Token Ring and Ethernet clients to communicate directly with the same server. This means that there is no need to populate high-speed servers with multiple NIC adapters to achieve the required bandwidth levels for client-server communications. Cisco has licensed the ISL technology to network interface card (NIC) vendors to produce high-speed, multi-VLAN NICs to provide network administrators increased design flexibility.
Why Cell-Based Backbones?
Second generation backbone switches provide full Token Ring and Ethernet ATM LANE clients, as well as LANE services. These features permit ATM backbones to offer similar design flexibility for network administrators who deploy ATM as for those who implement ISL backbone networks:
*Ethernet and Token Ring users can share the same physical ATM link.
*Ethernet and Token Ring users can directly communicate with the same ATM attached server. No translational bridging or routing is necessary.
*When communication between Ethernet and Token Ring clients is required, integrated layer three routing and translational bridging between the different media can be deployed. Alternatively, standalone routers also offer IP and IPX routing and translational bridging between Ethernet and Token Ring ELANs.
In addition, the redundancy of ATM networks is enhanced with support for Simple Server Redundancy Protocol (SSRP). SSRP provides the capability to configure redundant LANE services in the network in order to circumvent single point of failure scenarios. Hot Standby Router Protocol (HSRP) for transparent recovery of an IP gateway failure is an additional layer three routing feature that is available for enhanced network reliability.
Good News for End Users and Network Administrators
With the availability of second generation stackable and backbone switches, network managers can build a unified frame- or cell-based backbone for their Token Ring and Ethernet campus LAN traffic. ATM LANE provides a high speed, scalable, multi-VLAN cell-based solution, while ISL provides the equivalent high speed, scalable, multi-VLAN frame-based solution. Both backbone transport options work effectively whether the traffic is Ethernet, Token Ring, or a mixture of both.
Whether the goal for a network is to remain pure end-to-end Token Ring, pure Ethernet, or have the two coexist, legacy LAN investment can be protected. Future network migration can be accomplished at a rate that allows the company maximum financial leverage. In the meantime, by replacing shared media hubs with dedicated switch ports in the wiring closet, end users can enjoy switching performance to the desktop. Network administrators can enjoy complete flexibility in application deployment. And, network operation troubleshooters can enjoy enhanced problem determination, isolation and correction capabilities. The time has come for enterprises to deploy their switched infrastructures in the campus at a speed that makes business sense. Enterprises are no longer being held hostage by technological constraints.
ABOUT THE AUTHOR:
Eric Teagarden is Product Line Manager for Token Ring switching at Cisco Systems Inc. (Triangle Park, N.C.) and has been involved in Token Ring product development and network design since 1985 at Cisco Systems, Inc.and IBM. He can be reached at (919) 472-2650, via fax at (919) 472-2986, or e-mail at email@example.com.