The paper describes a low-level storage manager that is based on a prototype reduced instruction set computer. The aim is to simplify storage programming by means of a large virtual address space or a one-level-store, a database storage, hardware support for a transaction mechanism (similar to the use of hardware page faults), and a high-level programming language interface to all storage. The storage architecture provides segment registers to expand the virtual address space and an inverted page table for address translation.
Some performance data indicate that “the cost of database storage is not obviously too high” and that “the implementation cost is modest for hardware and software.” Surely, more experience is necessary to exploit the concept properly.
The paper is written very well and it gives much attention to implementation issues. I highly recommend the paper for all people interested in storage architectures, operating systems, or database systems.