rsion 1.0 permits only one LANE Server (LES) and Broadcast and Unknown Server (BUS) on an emulated network, creating the possibility of a much-dreaded single point of failure.
Life in the Fast LANE
LANE defines
how an Ethernet- or Token Ring-based network OS and applications run unmodified over an ATM network. LANE works by allowing the OS and all protocols at and above Layer 2 to seamlessly operate with ATM. An ATM adapter that uses LANE drivers -- based on Network Driver Interface Specification (NDIS) or Open Data-Link Interface (ODI) -- appears to the server's pr
otocol stack and NOS as an Ethernet or a Token Ring adapter. Address conversions are facilitated by implementing a client/server architecture whereby the media access control (MAC) addresses (Layer 2) are mapped to ATM addresses for all emulated end stations. Thus, existing LAN applications can be used on ATM networks without changing the underlying protocols.
LANE consists of four software components. The first three, all servers, are collectively known as LAN emulation services. Typically residing on ATM switches or edge devices (such as Ethernet or Token Ring switches with ATM uplinks), these services perform tasks such as indicating what type of LAN is being emulated; maintaining a table of all LANE end stations; and handling broadcasts, such as NetWare RIP and SAP messages and NetBIOS name queries. Individually, each performs a separate logical function.
The LANE Configuration Server (LECS) contains the configuration of the emulated networks. Typically it resides in a switch or a router and p
rovides configuration information to clients, including the ATM address of the LANE Server. This server is responsible for dynamically assigning LANE clients (LECs) to the different emulated LANs they desire, regardless of the physical switch port used.
The LES acts as a central clearinghouse for mapping between MAC addresses and ATM addresses. It can be configured as a dedicated station, integrated into a switch or edge device, or provided as software for a PC or workstation. It is typically implemented in the same device as the Broadcast and Unknown Server, discussed next. As noted, it also is the weak link in LANE because it's a potential single point of failure.
The BUS handles broadcast and multicast frames, as well as frames for which the destination unicast MAC address has not yet been resolved to an ATM address. It is a vital component because it resolves the inherent differences between traditional broadcast-based LANs, such as Token Ring and Ethernet, and the connection-oriented ATM netw
ork.
The fourth component, the LANE client, runs on every end station on the emulated LAN. Its main role is to initiate communications with the LECS and the LES and establish connections to other LECs. This includes requesting the LES to map MAC addresses to ATM addresses. Through the connections that are established, the LEC forwards data to the BUS and onto the target LEC. An LEC can be a workstation, file server, bridge, switch, or router. Each LEC has an ATM address and at least one MAC address.
Adding Fault Tolerance
Under version 1.0 of the LANE spec, hardware redundancy isn't possible on the emulated LAN. If the LES or the BUS fails, or is isolated from the network for any reason, then users cannot connect to the ATM backbone.
To address this single-point-of-failure problem, a number of independent vendors have sought to implement solutions that provide fault tolerance on an ATM network. For example, Cisco Systems recently announced its Simple Server Redundancy Prot
ocol (SSRP) that replicates the
configuration
information in the LANE Servers. SSRP adheres to the specs of the ATM Forum when communicating with the LEC, yet it eliminates the limitation of a single LECS or LES/BUS per emulated LAN. SSRP conveys information between the LECS and the LES/BUS devices as to which are operational and which are not. Basic configuration information about all emulated LANs is shared between the primary and up to three backup LECS devices.
If the LES/BUS fails, connections to the attached LECs are dropped. The client, sensing a network failure, attempts to reconnect to the emulated LAN. It will be directed by the LECS to a functioning backup LES/BUS on another network device. SSRP addresses the critical issue of ATM resilience by providing a software backup, enabling LANE clients to remain fully connected to the network.
SSRP, however, provides only part of the solution toward fault-tolerant LANE over ATM. The network may have a new fully functional
LES, but if a workstation is unable to discover this new location, the SSRP enhancement doesn't help. A second vendor, Olicom (for whom this author works), has completed a fault-tolerant solution on the LEC side that builds on SSRP and works within the ATM Forum 1.0 specification. This software (Dynamic Connection Redundancy, or DCR) enables a client to automatically search the ATM network for the location of the new LES/BUS through the LECS.
Working in tandem with SSRP, DCR does this by reading network configuration information via the Interim Local Management Interface (ILMI) from its ATM switch. The ILMI specification is an ATM Forum link management interface that enables two adjacent ATM devices to automatically configure the operating parameters of the common ATM link between them. This process, called
LECS discovery via ILMI
, lets the LANE client locate the new LECS with no need for manual reconfiguration. The usual alternative method -- the "well-known address," where the LECS is located
at a fixed address -- is limited because it does not normally provide for hot standby of these critical services and potentially results in far greater downtime for the ATM LANE network. While SSRP still provides for connection to a backup LECS even when the well-known address is used, DCR provides faster autoreconfiguration and recovery from a network failure. When this occurs, DCR performs a search and establishes a connection to the backup server automatically.
Hard Choices
For network managers who have shied away from ATM because of concerns about adding a nonredundant component to their otherwise fully fault-tolerant network, SSRP and DCR represent an early response to the problem of the single LES/BUS. Managers can deploy ATM with the confidence that they are covered by true, end-to-end fault tolerance. SSRP and DCR work within industry standards to provide for redundant LAN emulation services components, and they eliminate the single point of failure in existing LANE implement
ations. Although the ATM Forum Technical Working Group expects resolution of the single LES/BUS problem in version 2.0, for a network manager, six to 12 months is a long time to wait.
illustration_link (28 Kbytes)
LAN emulation (LANE) maps Ethernet or Token Ring addresses to ATM addresses.
illustration_link (35 Kbytes)
Simple Server Redundancy Protocol and Dynamic Connection Redundancy reestablish logical links to devices on the ATM segment.
Steen B. Lohse (
slo@olicom.com
) is vice president of marketing at Olicom. He holds degrees in engineering and business administration from the University of Copenhagen.
Reliable ATM Networking
