No abstract

has been involved in educational research and evaluation with a focus on science, technology, engineering and mathematics (STEM) education and issues of race/ethnicity, gender and disability since the mid 1970's. Dr. Campbell, formerly a professor of research, measurement and statistics at Georgia State University, has authored more than 100 publications.

APCAM was designed as a practical intelligent parallel search disk for support of relational data base systems. The aim was not to provide a complete relational engine but to provide a machine which would search large volumes of data and perform projections and restrictions along with helping with sorting. A device with these somewhat limited objectives would provide a major part of the performance necessary to make large relational data base management systems practical by providing a cost-effective adjunct to existing hardware and software. The interesting part of the design is the effort to solve certain real world problems which the more research oriented designs have ignored. That is, the design was to be built with components and techniques that were well within the state-of-the-art.Thus, while the inspiration for APCAM came from such research projects as CASSM (Su, 1973(Su, , 1975, P~AP (Ozkarahan, 1975), and the Data Base Computer (Baum, 1976, it was felt that all these projects had overlooked certain aspects of the real world.For example, for large data bases moving-arm disks are the only cost-effective storage media. Fixed head disks are too expensive to store large data bases and staging from moving arm to fixeddisk does not solve the real problem of access to large volumes of data. Further, the write-afterread head design of CASSM is really not possible since it is difficult enough to position a single head over a track, let alone two heads. More importantly, read heads of sufficient sensitivity to read current disks would be completely overwhelmed with a write-head nearby.With modern disks comes higher density and with higher densities also comes higher error rates --both single errors and burst errors. Schemes which rely on marks on the disk, read or written on the fly, become unworkable since on a modern disk one can only be certain of a given bit after the entire block has been read and the error detection and correction applied if necessary. Thus, it is necessary thatthe design be block oriented and buffered such that decisions can be made or remade on the basis of corrected data.Further, while APCAM shares much, at least philosophically, with the Mass r~mory of Hsiao's Data Base C~mputer (DBC), the DBC concept as a whole is impractically complex --particularly in terms of the number of different types of logic units and storage devices. The Colm~ands APCAM places sufficient intelligence and search logic in each movable-head disk drive so that one entire cylinder can be searched in one revolution to retrieve the desired records or fields of records. Thus, each cylinder on a drive (or multiple cylinders on multiple drives) can be considered as a cell, all of whose records can be addressed by content in one unit time (namely a revolution).The environment envisioned for APCAM is that of large data bases, multiple large disk drives, where the number of records retrieved, changed or deleted at any one time is small. Thus, the entire data base can be searched but only a few percent of the records would...

No abstract

APCAM was designed as a practical intelligent parallel search disk for support of relational data base systems. The aim was not to provide a complete relational engine but to provide a machine which would search large volumes of data and perform projections and restrictions along with helping with sorting. A device with these somewhat limited objectives would provide a major part of the performance necessary to make large relational data base management systems practical by providing a cost-effective adjunct to existing hardware and software. The interesting part of the design is the effort to solve certain real world problems which the more research oriented designs have ignored. That is, the design was to be built with components and techniques that were well within the state-of-the-art.Thus, while the inspiration for APCAM came from such research projects as CASSM (Su, 1973(Su, , 1975, P~AP (Ozkarahan, 1975), and the Data Base Computer (Baum, 1976, it was felt that all these projects had overlooked certain aspects of the real world.For example, for large data bases moving-arm disks are the only cost-effective storage media. Fixed head disks are too expensive to store large data bases and staging from moving arm to fixeddisk does not solve the real problem of access to large volumes of data. Further, the write-afterread head design of CASSM is really not possible since it is difficult enough to position a single head over a track, let alone two heads. More importantly, read heads of sufficient sensitivity to read current disks would be completely overwhelmed with a write-head nearby.With modern disks comes higher density and with higher densities also comes higher error rates --both single errors and burst errors. Schemes which rely on marks on the disk, read or written on the fly, become unworkable since on a modern disk one can only be certain of a given bit after the entire block has been read and the error detection and correction applied if necessary. Thus, it is necessary thatthe design be block oriented and buffered such that decisions can be made or remade on the basis of corrected data.Further, while APCAM shares much, at least philosophically, with the Mass r~mory of Hsiao's Data Base C~mputer (DBC), the DBC concept as a whole is impractically complex --particularly in terms of the number of different types of logic units and storage devices. The Colm~ands APCAM places sufficient intelligence and search logic in each movable-head disk drive so that one entire cylinder can be searched in one revolution to retrieve the desired records or fields of records. Thus, each cylinder on a drive (or multiple cylinders on multiple drives) can be considered as a cell, all of whose records can be addressed by content in one unit time (namely a revolution).The environment envisioned for APCAM is that of large data bases, multiple large disk drives, where the number of records retrieved, changed or deleted at any one time is small. Thus, the entire data base can be searched but only a few percent of the records would...