Overhead efficiency is a critical topic for SPs that charge for the amount of bandwidth capacity customers use. ATM was introduced to the SP market as the technology that would enable converged voice, video, and data traffic to reside on the same infrastructure (because of the intrinsic QoS parameters built in to the technology). The technology is widely used by Internet service providers (ISPs) today, but many of the original intentions behind ATM have not been used due to the complexity of configuring, selling, and maintaining such features. ISPs want to maximize their profits and minimize the costs associated with transporting IP over ATM.
ATM uses fixed-sized ATM cells of 53 bytes. Each cell's composition includes 5 bytes of fixed overhead and 48 bytes of data. Depending on the ATM adaptation layer (AAL) used, the amount of AAL overhead can be as high as 4 bytes in addition to the 5 bytes of fixed overhead. This extra AAL overhead could result in as little as 44 usable bytes in a 53-byte cell. This equates to an approximate efficiency level of 83 percent (or 17-percent overhead). SONET's TOH and POH combined equal 36 bytes per the calculation that follows:
TOH (Transport Overhead)
9 rows _ 3 columns = 27 bytes
POH (Path Overhead)
9 rows _ 1 column = 9 bytes
TOH + POH = 36 bytes
The 36 bytes of overhead used in SONET represent approximately 4 percent of the total 810-byte STS frame. The ATM inefficiencies are further compounded when the variability of data sizes is calculated. Because of the nature of web pages, most Internet browsing traffic consists of many small-size packets. Most traffic that is generated originates from workstations connected to a LAN with Ethernet technology. The smallest PDU available in Ethernet networks is 46 bytes but can vary up to the maximum transmission unit (MTU) size of 1500 bytes. If a packet does not fall neatly on a cell boundary, the rest of the cell is padded.
In the case of a frame sent with a frame size of 64 bytes, two ATM cells are needed to transport the data. The first cell would be fully used at 48 bytes of payload (assuming that an AAL with no extra overhead is in use), and the second cell would be nearly empty with only 16 bytes of payload. This scenario results in a low efficiency level (approximately 43 percent). Figure 9-8 illustrates the PoS efficiency over ATM in both a line graph and table, which compares efficiency based on packet size.
Figure 9-8 PoS Efficiencies Compared to ATM