of the disk as unreliable or unusable. In this case, the master file is able to know that certain parts of the disk should no longer be used.
In addition to the performance aspects of the master file, the master file concept can also be used to undelete files. When a file is removed or deleted from a disk drive, the only action that generally happens is that the data in the master file about the deleted file is removed from the master file. However, the physical region (or regions) of the spinning disk that actually stored the data is unchanged. Obviously, if the information in the master file is not completely removed during the delete process, but simply put aside for future use, it would be possible to restore the file that was deleted. Of course, if the physical region of the disk (which would then be considered available for other files) is re-used for new data by the master file, then the “undelete” process is not going to work very well. In general, PCs have had the undelete capability for many years; however, the work- stations have never had this capability. One reason not to include this capability is
Computer Hardware Basics 21
the potential for slowing down the overall speed of data access. If the master file completely “forgets” the deleted file information, then the master file can remain more compact.
Another issue with respect to disk drive behavior is the importance of how the data in a computer file are actually located on the spinning disk. This is espe- cially true for the very large files, and CAD files, particularly 3-D models, tend to be rather large (as mentioned earlier with respect to the memory system it could be 100s of megabytes). With these large files the data is going to far exceed the capacity for storage of one particular region of the disk (that is a particular cylin- der or block). In this case, the data is going to be written to a large portion of these regions, and the master file is going to have to keep track of where all those regions are located.
Clearly it would be best if all that data is written to regions of the disk that are “next to each other” or contiguous. In this case, the master file will only have a small amount of information to store; i.e., where the large file starts on disk and how far to go to read all the data. On the other hand, if the large amount of data in a big CAD file is broken into many small pieces, and these small pieces are in regions of the disk that are not contiguous, then the heads above the spinning disk have to move many times (perhaps 100s or even 1000s of times) to reassemble the entire file into the memory system. Clearly this situation is going to be pretty devastating to the performance of the disk drive and thus the entire storage sys- tem. This situation is sometimes called “thrashing,” although this term may be best considered to apply to the situation of the memory system having to unload data to the storage system, and the storage system has to rapidly attempt to shut- tle data to and from memory (or swapping).
One important method to prevent this breakdown in performance of the disk drive is to not let the disk drive capacity become too near to its maximum. As the used regions of the disk grow, the available regions not only become smaller (since less and less free space is available), these regions become spread apart as the larger files create smaller gaps. As a rule of thumb, no disk drive should exceed 80 percent of its maximum capacity. On the other hand, if a disk drive is 90 or 95 percent used, one can be certain that performance is suffering as the heads attempt to find and assemble files.
It was mentioned earlier that reformatting a disk drive will improve perfor- mance, and indeed, this will help alleviate the situation of the heads not getting data efficiently due to the poor organization of the data on the disk. However, it was also mentioned that the reformatting will erase all the data in the disk drive.