* The bottom line of the frames vs. cells debate
In the last newsletter, we determined that the historical arguments as to why frames are inherently better than cells are quite weak in today’s environment. In this newsletter, we’ll see if any of the pro-cell arguments have survived.
In the last newsletter, we determined that the historical arguments as to why frames are inherently better than cells are quite weak in today’s environment. In this newsletter, we’ll see if any of the pro-cell arguments have survived.
Historically, ATM, as the premier cell-based technology, has been positioned as being better than frame technologies because the cells are easier to process and latency can be controlled more readily.
Let’s start with the processing issue. It’s quite true that 10 to 15 years ago, the processing power necessary for dealing with cells vs. frames was a major issue. Due to the simplicity of the packet format, it’s a given that cells are indeed simpler to process than frames. However, with Moore’s Law continuing to hold, processing power in networking equipment is no longer a major issue. Further, as pointed out a couple of weeks ago, even though the equipment itself may have an internal cell-based architecture, the majority of the interfaces to this equipment – and consequently the service interfaces that customers see – are frame-based.
Now let’s move to latency. The argument has been that latency is easier to control in a cell-based environment. The real reason is because cells tend to be shorter than frames. The key to controlling latency is the “packet time,” which we define as the number of bits in a packet divided by the transmission speed. As the transmission speed increases, the “packet time” for a given packet length decreases. Putting some real numbers around this, the “packet time” for an ATM cell transmitted at 56K bit/sec is roughly the same as the time for a maximum-size frame relay frame at T-1/E-1 speeds. (And most frames never approach the maximum frame size.)
The end-result is that while cells are fine and indeed have very low latency, the increase in normal transmission speeds has obviated this as a “must-have” reason for ATM.
So what’s the bottom line from this discussion? You should be very careful about making any blanket assumptions about frames vs. cells based on decade-old arguments. In this case, both technologies work fine and there’s no strong inherent advantage to one over the other.
