JD

When I started programming punched cards were commonly used. Cards you say?! What does this have to do with disks as RAM?
If you are asking that question then apparently you never used cards, and therefore will not establish a relationship. A little explanation might be required. A card based dataset begins life (in this example case) at remote locations where data is collected and punched onto cards. The cards may then experience a pre-sort and classification process. Then the cards are (were) shipped by mail, common carrier, or courier. From data sampling to first shipment might take weeks or months (and some time years). Once the cards get to the data processing center they get classified again, may get aggregated with cards from other sources, the card boxes themselves may get sorted (e.g. by box zip code), and placed into priority queues for processing. All these priority queues of card decks are passed through one or several card readers and are either directly used by the processor or transcribed onto multiple faster medium.

The above process is a type of process that lends itself well for penalization. Data is collected from dispersed locations and is organized into a fast flowing stream for processing. This particular problem does not require random access across the entire length of the data stream. Problem solving is limited to a small portion of the data stream as it passes through the program. This is a classic example of a pipeline.

If a given problem lends itself to a pipeline solution method then, yes, disks (or other storage media) in conjunction with the means of consolidating many moderately fast sequential devices into one or a few very fast pipelines, then in this case, the disks can be thought of as RAM.

For problems where the Random Access portion of the problem can be performed in parallel far away (time or distance) from the high-speed sequential portion of the problem then a throughput analysis can determine the number of servers (systems with disks), and network infrastructure necessary to keep the pipeline from running dry.

Jim Dempsey

And then I realized that the webpage was from back in 2003. In five years, we've seen the price slashed from $150/GB to $10/GB and the size has come down from a 5.25" disk bay to my thumb. Of course, my drive isn't configured for RAID.