Parallel SCSI

Parallel SCSI (formally, SCSI Parallel Interface, or SPI) is the earliest of the interface implementations in the SCSI family. In addition to being a data bus, SPI is a parallel electrical bus: There is one set of electrical connections stretching from one end of the SCSI bus to the other. A SCSI device attaches to the bus but does not interrupt it. Both ends of the bus must be terminated.

SCSI is an intelligent, peripheral, buffered, peer-to-peer interface, hiding the complexity of the physical format. Every device attaches to the SCSI bus in a similar manner. Up to 8 or 16 devices can be attached to a single bus. There can be any number of hosts and peripheral devices but there should be at least one host. SCSI uses handshake signals between devices, SCSI-1, SCSI-2 have the option of parity error checking. Starting with SCSI-U160 (part of SCSI-3) all commands and data are error checked by a CRC32 checksum.

The SCSI protocol defines communication from host to host, host to a peripheral device, peripheral device to a peripheral device. However most peripheral devices are exclusively SCSI targets, incapable of acting as SCSI initiators—unable to initiate SCSI transactions themselves. Therefore, peripheral-to-peripheral communications are uncommon, but possible in most SCSI applications. The Symbios Logic 53C810 chip is an example of a PCI host interface that can act as a SCSI target.

The first two formal SCSI standards, SCSI-1 and SCSI-2, included parallel SCSI as a central part of the protocol. The SCSI-3 standard then split the framework into separate layers so parallel SCSI is now just one of a number of available implementations. See the main SCSI article for a complete list. As with all types of SCSI bus, parallel SCSI communication takes place between an initiator and a target.

...
Wikipedia