4-5 Inverse Multiplexing ATM, Bit by Bit
ROBIN D. LANGDON
Not long ago, the deployment of asynchronous transfer mode (ATM) was considered a wildfire marketATM would be available on the desktop, would be used in LAN backbones, and would provide transparent LAN-to-WAN interconnection at almost infinitely scalable bandwidths with all the benefits that classes of service could provide. However, the enthusiasm for ATM might be judged to be on the wane.
Completely dismissing ATM would be premature. ATM might be having trouble competing with switched and fast Ethernet for the desktop, but it has firmly established a home in corporate and carrier backbones, where it is experiencing strong growth. On the access side of the public arena, however, ATMs migration into the WAN has been and is projected to be slow. Based on industry forecasts, ATM service revenues in the public network will remain relatively small when compared to alternative switched data network services such as frame relay. A new access option, inverse multiplexing over ATM, is becoming a reality for corporate users who want the benefits of ATM bandwidth but want to avoid the high costs of ATM service and implementation.
ATMS POSITION IN THE MARKETPLACE
For end users and carriers, the real issue is what to do when the local network needs to expand to the enterprise network and runs into the bottleneck of the public WANs access bandwidth. ATM outside the backbone and in the wide area is a possible solution, if ATM had a larger presence in the public network. ATM service is by no means ubiquitous, nor is it expected to become so in the next several years. ATM in the WAN may not even be available for those customers who need it or are willing to pay for it.
Some network managers, even those running their corporate backbones at ATMs OC-3c rates, may not be able to justify the steep price tags (relative to access alternatives) or backhauling expenses associated with OC-3c or DS3 WAN connections. Although T1 ATM prices are coming down to the point where they are equal to or less than the prices of traditional T1 lines (especially in areas where the carriers are trying to encourage users to experiment with ATM services), the bandwidth lost to ATM cell overhead and partially filled cells reduces the available bandwidth to the point where the inefficiencies of T1 ATM may be too high a burden for an application to bear.
To complicate ATMs position in the marketplace, it is also faced with formidable competition from frame relaya service that has no cell overhead, is ubiquitous, is priced attractively, and has tremendous market momentum. Like ATM in the LAN, ATM in the WAN must compete against alternative technologies that are less expensive, readily available, or easier to use.
INVERSE MULTIPLEXING FOR ATM
T1 inverse multiplexing, or imuxing, is a process in which a single data stream is split across multiple T1 lines in a round-robin fashion; the T1s are logically combined to form a single virtual data channel that is the aggregate of the T1 bandwidths (minus a small amount for overhead). From the point of view of the device providing the data stream, it is communicating via a single, high-speed WAN channel at some multiple of the T1 ratethat is, at the bandwidth of a fractional T3 service, but using readily available, less expensive T1 services.
The similarities between inverse multiplexing and ATM are significant when planning for current and future network implementations. ATM and inverse multiplexing topologies can both provide the ability to link individual sites by clear channel broadband data pipes. Both imuxing and ATM provide scalability, and both seamlessly link LANs and WANs in enterprise networks. Inverse multiplexing and ATM complement each other and can work together hand-in-hand.
Where the rate of a traditional T1 is insufficient and T3 is too expensive or is unavailable, T1 inverse multiplexing is an efficient and immediate cost-effective solution to provide increased bandwidth. The concept of inverse multiplexing can be applied to ATM cells, where an ATM cell stream is transmitted across multiple T1/E1 links; alternatively, it can also allow ATM cell traffic to be transported across the existing T1/E1 network infrastructure as a bitstream. In either case, users avoid having to pay for the excessive bandwidth of a T3 or OC-3c line that they may not need. The customer also avoids the price, in dollars or bandwidth, of T1 ATM service.
There are two variations of inverse multiplexing for ATM: cell-based and bit-based. Each has its own strengths, discussed at greater length in the following sections.
|