Upgrading fddi backbones to atm
Managers whose bandwidth demands have outstripped the capacity of their fddi networks have two upgrade choices. The right pick can save both time and effort.
Ron Shani Hynex
For the past several years, Fiber Distributed Data Interface (fddi) has been a popular choice for mission-critical backbones because of its dual-redundant 100-Mbit/sec fiber-optic structure, inherent fault tolerance, long-distance coverage, and maturity. Yet many network managers today find that their backbone bandwidth demands have surged beyond the capacity of fddi, so they must look for a way to increase their backbone bandwidth. For those seeking a proven, tested, and immediately available solution, there are two choices: an fddi switch or an Asynchronous Transfer Mode (atm) backbone.
An fddi switch is a device that provides fast switching or bridging among its many fddi ports. While the use of an fddi switch will increase overall network capacity, it limits the network`s collapsed backbone to a single location, as the network remains operating at 100 Mbits/sec with no option to move to a higher-speed backbone.
Using an atm super-backbone with devices that connect each fddi segment to atm provides a better solution. This is because it switches between the many fddi segment rings--similar to the way the fddi switch does--but it also adds the ability to further expand the network and increase the backbone speed to 622 Mbits/sec and even to the gigabit-per-second range. This article will describe how the evolution from fddi to atm can be accomplished.
How to upgrade?
Two ways to upgrade an fddi network to atm are "forklift" and "step-wise." A forklift upgrade appears very simple: Buy the new equipment, including an atm switch, Ethernet-to-atm switch/ bridges, and a network management system; allocate enough time; and then start to work. In practice, however, forklift upgrades can prove quite complicated. They require that technicians take down the fddi backbone, replace all the fddi-to-Ethernet bridge/switches with atm-to-Ethernet devices, install the atm switch, connect the cables, and power up the units. Then they must configure each new unit and hope that the network will come alive.
This process is obviously very cumbersome because it involves taking down the backbone and disrupting the organization`s normal operations. Since mission-critical data flows through the fddi backbone, no one can continue to operate while this upgrade is taking place. As a result, the networking staff will have to perform the upgrade during the organization`s off time, perhaps during a weekend or holiday, and hope that by the time the organization resumes normal operation the new network will be up and running.
Is this the best way to upgrade a vital network? Certainly not: No one can guarantee that the network will be fully functional on time. Even if the networking staff completes its tasks on time, it will not be possible to guarantee that the network will be tuned and will function properly.
Conversely, the step-wise approach enables migration to atm one step at a time. Each modification to the network is minimal and easy to perform and manage. Small steps also enable downgrading if problems arise during the upgrade process. By making minimal changes to the existing network, the initial investment will also be minimal. To upgrade the first part of the network, only basic new equipment is required. Once the networking staff becomes familiar with that technology, they can continue with the upgrade process--one step at a time.
Let`s follow an optimal upgrade process from an fddi-based network to an atm super-backbone. Figure 1 shows a typical fddi network, where some servers connect directly to the fddi backbone for high-speed access, and many clients connect to the backbone through Ethernet-to-fddi bridge/switch devices. In many applications there may be more than one fddi ring connected through a router or fddi switch, but for simplicity, this example will show just one ring.
The first step in the process will be to connect the fddi-to-atm edge device to the existing fddi backbone, connect its atm port to an atm switch, and connect a PC with an atm network interface card to another port of the atm switch (see Fig. 2).
This initial step can be performed while the backbone is up and running, with no service disruption. The first step establishes a small, functioning atm network that connects to the fddi backbone for testing and familiarization with atm technology. Once the fddi backbone is connected to a small atm network, the networking staff can take the time to configure each atm device, run applications from the PC through the atm connection, and verify that everything functions properly.
Once the staff members feel comfortable with the new atm technology they can move to the second step. Here, the staff connects another fddi-to-atm node to the fddi ring and to the atm switch (see Fig. 3).
Before connecting the second fddi-to- atm edge device, the staff will need to configure the Spanning Tree weight parameter so that it will be the first unit to disconnect. In this step the staff tests the stability of the atm network as well as the functionality of the fddi-to-atm switches, the atm switch, and any application software chosen.
The staff should continue to test the network, noting the Spanning Tree messages that flow through this small atm network. One can test application software end-to-end using the atm-attached PC and any of the fddi-attached stations (directly or through Ethernet).
Once these tests are completed and everything functions as it should, the staff proceeds to the third step. Here, actual modifications to the fddi backbone are made. This step calls for opening the ring and segmenting it into two operational rings (see Fig. 4). Once this reconfiguration is completed, the two fddi rings will communicate through the atm switch.
The Spanning Tree now detects that there are no longer any loops and will instruct the fddi-to-atm edge devices to assume a forwarding mode. It will take a few seconds, and then the network will continue to function at double the capacity, at 200 Mbits/sec.
The staff should test the new topology to see if everything continues to function. If not, they can start to troubleshoot by first looking at the new device`s configuration. If this takes too long they can reverse the last step, reconnect the fddi ring to one long segment, and troubleshoot at their convenience, without the pressure of users not being able to communicate.
Once the new network is functioning properly, the staff can take some time to gain confidence in the new technology now that it serves the mission-critical applications. There`s no need to hurry--in this step-wise upgrade approach each step is minimal and reversible.
Now that the atm and fddi networks are in place and the staff members are content with the results, they can proceed to the next step: expanding the atm super-backbone. There are many ways to grow the network, and an organization can choose the one that best fits its network needs. It can
extend the network to the atm side
through atm-to-Ethernet/Fast Ethernet access units,
further segment the fddi ring into
three or four smaller networks,
add redundancy to the new atm
backbone.
Figures 5 through 7 show examples of networks that expand based on the above options. The first option, illustrated in Fig. 5, calls for adding an atm-to-Ethernet access switch to connect Ethernet-based stations directly to atm. The second option, illustrated in Fig. 6, calls for adding another fddi-to-atm edge device to segment the fddi ring into three smaller rings, while gaining more bandwidth. The third option, redundancy, increases the reliability and uptime of the new networks (see Fig. 7).
A few words on redundancy are in order. fddi networks are very reliable and robust. They include a dual counter-rotating ring, so that if one ring fails the network resumes operation on the secondary ring. Moreover, if a station fails, the other knows how to bypass it. This robustness is a crucial element in building a mission-critical organization backbone, and it would be essential to have the new atm backbone function similarly.
Unfortunately, atm networks do not have inherent redundancy, and it is up to the equipment vendors to implement it. However, only a few vendors provide this important feature. Redundancy in atm networks requires two functions: a dual physical link (two ports with internal switchover) and local area network (lan) emulation client fail-over support. Any atm device that uses lan emulation (see "Interoperability of atm and lans" on this page) connects to several servers such as bus and lan emulation configuration server. When switched to another connection, the device should automatically search for the atm servers and log into them. Alternatively, the atm switch may feature private network-to-network interface support, which can route the redundant fddi-to-atm edge device connection to the appropriate services.
It is recommended that each fddi-to-atm edge device in use include a dual atm uplink to maintain fddi`s non-stop service.
Summary
In this article we examined the ways to upgrade an fddi-based backbone to the new and faster atm technology. We showed two upgrade options--forklift and step-wise--and explain why step-wise is the best approach. We further emphasized the need to use redundant atm connection and whatever features are necessary to achieve usable redundancy. u
Ron Shani is director, business development of Hynex, a subsidiary of Elbit Ltd., based in Israel. His expertise in local-area and wide-area networks has included participation in the ieee 802 standard committees and the atm Forum. He can be reached at [email protected].
Interoperability of atm and lans
Asynchronous Transfer Mode (atm) and local area networks (lans) originated from different perspectives. atm came as a follow-on to Integrated Services Digital Network (isdn) as "wideband" services for isdn and originated in telephone company standards committees. lans, on the other hand, were developed by the computer industry. These different perspectives resulted in different approaches in the design foundation of each technology. atm resembles the telephone-centric network, while lans use distributed and fair networks. The major technical difference between atm and lans is that atm uses connection-oriented networks and lans use connectionless networks.
When two stations need to exchange data in an atm network, they first have to establish connection, set up connection parameters, and then transfer the data. In lan networks the source station will simply transmit the data to the network, without any knowledge of the destination location, and the network is responsible for delivering the data to its destination.
One of the first tasks faced by the designers of atm technology was to establish connectivity between atm and lan networks. This effort led to the "lan Emulation" recommendations of the atm Forum. The lan Emulation standard (lane) establishes the way lans connect to and over atm networks. It defines, among other things, the way lan frames are sliced into atm cells and then assembled at the receiving end; the way connection is established over the atm network such that the source and the destination stations remain unchanged; and the way broadcast and multicast messages are transmitted over the atm network.
The lane standard defines two encapsulation formats for lan data: Ethernet and Token Ring. Each of the formats defines parameters that are specific to the lan source, such as frame length (1500 bytes in Ethernet and 17 kbytes in Token Ring) and frame structure. As a result, stations that implement Ethernet lane do not communicate with stations that implement Token Ring lane.
Where does Fiber Distributed Data Interface (fddi) stand in all this? There is no fddi lane standard. The designers of the lane standard left it open, recommending that fddi be implemented as a Token Ring lane due to the similarity between fddi and Token Ring. Doing so, however, there will be no connectivity between an fddi ring and the stations that attach to it and the atm stations that implement Ethernet lane.
Most networks that use fddi as a backbone use it to connect Ethernet and even Fast Ethernet stations among themselves and with fddi-attached stations. There are ieee standards that define how to convert Ethernet frame format to fddi, such that all stations can communicate (Translation standard, 802.1h and i).
When upgrading the fddi backbone to atm, the network design has to preserve end-to-end connection and communication. Therefore, networks that used to connect Ethernet segments to fddi, and plan to expand by connecting future Ethernet and Fast Ethernet stations to atm, need to ensure that the fddi-to-atm access units perform Ethernet lane, long-frame process/Internet protocol fragmentation, and fddi translation. Only fddi-to-atm access devices that implement these processes will ensure interoperability between the existing fddi/Ethernet network and the new atm network.