Frame vs. cell: 10 years on

Last time we began a discussion of the historical impact and current status of frame- vs. cell-based communications architectures and reviewed some of the key technical issues. Most of the reasons that the proponents on each side of the issue gave for why their approach was “better” 10 years ago don’t hold up today.

Let’s start with the argument that cells are easier to process.  This is true, but it only matters if you don’t have enough processing power.  Today’s generation of switches has sufficient processing power to switch frames or cells at gigabit speeds.

There was also the latency argument that, especially for multimedia applications, very small packets were needed to provide low delay and excellent traffic management. Two problems here: First, the small packets and maximum packet size issues were always muddled.  Second, this is really only a significant problem for low speeds like 56K/64K bit/sec. At T-1 speeds and above, most of these problems disappear anyway.

The frame arguments likewise are evaporating.  One of the best pieces of frame relay marketing in the past decade was the discussion of ATM’s “cell tax.”  This referred to the additional overhead incurred for ATM as opposed to frame relay when multiple packets are transmitted instead of a single packet.  But this argument falls apart for the same reason as the latency argument.  Typical services are much faster and much cheaper than they were a decade ago.  Consequently, squeezing every possible bit-per-second out of the transmission facilities is less important.

So the bottom line is that while there may be a few technical reasons that you might prefer one technology vs. the other, these arguments will most likely be overshadowed by economic considerations.  Go for the best deal – not the “best” technology.