Interfaces between parallel optical memories and electronic computers

Abstract

Three-dimensional optical memories are considered in the context of information processing architecture with the objective of identifying their suitability for general and special-purpose computation. Their suitability is better delineated by the memories' limitations rather than their abilities. Optical memories have the potential to offer inexpensive and fast access to large sets of data. Furthermore, they are immune to electromagnetic interference and they exhibit fault-tolerance that is unusual in other types of storage including magnetic disks and capacitive solid-state devices. Data recording in optical memories is a rather slow process. This fact renders optical memories more suitable for archival storage application, where large storage volumes are required and high data densities are desirable. The question for a suitable optoelectronic interface emerges next. An optoelectronic interface for three-dimensional optical memories should meet certain design specifications. These specifications are derived from many considerations including data error minimization, performance capacity, hardware complexity, etc. Data error minimization, specifically reduction of intersymbol interference at the interface regime led to some geometrical design specifications. These specifications are consistent with regular patterns that reduce hardware complexity and enable system scalability. Performance capacity, in the context of available electronic computer technology, has also led to design specifications regarding the size of data pages and timing of interface's functions. Again, these specifications are scalable to the performance characteristics of modem electronic computers. Finally, architectural and functional considerations led to the investigation of algorithms for improved performance in data processing. We demonstrated a functioning auto-associative processor and we proposed a fast class of algorithms for data processing in bit-slice mode. Our findings support the notion that optical memories do have the potential for becoming a leading technology in archiving applications. It may also be possible that with advances in recording techniques, optical memories may one day be suitable for secondary storage, random access, and cache applications. When this occurs, optical memories will contribute to the flattening of the conventional memory hierarchy generally accepted today. But in their current form, optical memories matched with a suitable interface may be ideal architectures for applications involving associative access to data.