DBC 1012
The DBC 1012 Data Base Computer was introduced by Teradata Corporation in the early 1980s, as a back-end data base management system for mainframe computers. The DBC 1012 harnessed multiple Intel microprocessors, each with its own dedicated disk drive, by interconnecting them with the YNET switching network in a massively parallel system. The DBC 1012 was designed to manage databases up to one Terabyte (1,000,000,000,000 characters) in size.
The major components of the DBC 1012 architecture included:
- Mainframe Resident Software to manage users and transfer data
- Interface Processor (IFP) - the hardware connection between the mainframe and the DBC 1012
- YNET - a purpose-built system interconnect that supported broadcast and sorting
- Access Module Processor (AMP) - the unit of parallelism: includes microprocessor, disk drive, file system, and database software
- System Console and Printer
- TEQUEL (TEradata QUEry Language) - an extension of SQL
The DBC 1012 was designed to scale up to 1024 YNET interconnected processor-disk units. Rows of a relation (table) were distributed by hashing on the primary index.
The DBC 1012 used a 474 megabyte Winchester disk drive with an average seek time of 18 milliseconds. The disk drive was capable of transferring data at 1.9 MB/s although in practice the sustainable data rate was lower because the IO pattern tended towards random access and transfer lengths of 8 to 12 kilobytes.
The processor cabinet was 60 inches high and 27 inches wide, weighed 450 pounds, and held up to 8 microprocessor units.
The storage cabinet was 60 inches high and 27 inches wide, weighed 625 pounds, and held up to 4 disk storage units.
The DBC 1012 preceded the advent of RAID technology, so data protection was provided by the FALLBACK feature, which kept a logical copy of rows of a relation on different AMPs. The collection of AMPs that provided this protection for each other was called a cluster. A cluster could have from 2 to 16 amps.