Design and Implementation of a Minicomputer Catalog System.

Peter John Zombori · 1980

Under the PDP-11/70 minicomputer operating system a user must specify the unit upon which a mass-storage file resides. This procedure leads to inefficient use of storage space. One drive can be returning disk full error messages while other disks remain relatively empty. A more efficient use of available massstorage space would be beneficial. The following catalog system was found to best fill the above requirement. The user supplies the filename of the file to be saved and the cataloger returns the number of the unit which contains the most free storage space. The cataloger also supplies the location of files already on and deletes catalog entries for files which no longer exist. It proved desirable to have catalog capabilities in both batch and interactive environments. In an attempt to minimize core storage, storage, access time and cataloger complexity, the following file entry access scheme is implemented. Catalog file entries are kept on a primary and a secondary file. The primary file is accessed in a random mode by hashing the file name of the entry. If the file is not foundat the hashed address or up to a number of entries following the hashed address, then a linear search is made of the secondary file. With proper primary file size and a good hashing technique the secondary file should be relatively short and its use infrequent. The above procedure permits a rapid confirmation whether a file entry exists. Most other techniques require time consuming linked lists or exhaustive searches. The above access scheme is designed to respond within SGF timing requirements for user batch jobs. However, maintenance and configuration control requirements require an interactive list capability. The interactive catalog list option has a fairly slow response time. The list option lists all files in a user defined category. This requires an exhaustive search of the catalog files. The cataloger is designed to permit batch and interactive list commands to run concurrently by sharing access to the catalog files. Thus batch response time is not degraded by interactive list command processing. In summary, the cataloger is a set of structured assembly language programs which allocate, list and delete catalog entries for batch and interactive users. The cataloger is designed and implemented with special considerations for access time, storage requirements and operating environment.

Read the paper · More papers on PaperTik