Files are out, web services are in

The trusty file has been with us (that is, us PC users) since before the dawn of time. (If we consider the "dawn of time" to be the introduction of the IBM PC.) It has been the workhorse of data containers, going back to the elder days of CP/M and DEC PDP-11 computers.

While files have served us well, they are poor citizens of cloud computing. Files insist on being, on existing, at some location (usually on a disk). Because they have a known physical location, they are problematic for the cloud. A file is a singular thing, existing on a singular host. Even files on a device such as a SAN are on a singular host -- the SAN, arguably, is a host. A single point of access is a bottleneck and a threat to performance.

But it turns out that we don't need files. Files are nothing more that a collection of bytes. Some consider them a stream. Others will think of them as a collection of not bytes but characters.

We need collections (or streams) of bytes (or characters). But we don't have to house those collections in files.

In the cloud, we can use web services. Web services can provide collections of bytes (or characters) just like files. Web services can be distributed across multiple hosts, eliminating the bottleneck of files. Unlike files, web services are identified by URIs, and this gives us flexibility. A "file" provided by a URI can come from one server or from one of many (although you must ensure that all servers agree upon the contents of the file).

The design of a classic PC application involved the specification of input and output files. The design of a cloud application must involve the specification of web services.