 | |||
| Type | Bus | ||
|---|---|---|---|
| Production history | |||
| Designer | Team led by Larry Boucher atShugart Associates | ||
| Designed | September 1981; 43 years ago (1981-09)[1] | ||
| Standardized | ANSI X3.131-1986, June 1986 | ||
| Produced | Since 1983 | ||
Small Computer System Interface (SCSI,/ˈskʌzi/SKUZ-ee)[2] is a set of standards for physically connecting and transferring data between computers andperipheral devices, best known for its use with storage devices such ashard disk drives. SCSI was introduced in the 1980s and has seen widespread use on servers and high-end workstations, with new SCSI standards being published as recently asSAS-4 in 2017.
The SCSI standards definecommands, protocols, electrical, optical and logicalinterfaces. The SCSI standard defines command sets for specific peripheral device types; the presence of "unknown" as one of these types means that in theory it can be used as an interface to almost any device, but the standard is highly pragmatic and addressed toward commercial requirements. The initialParallel SCSI was most commonly used forhard disk drives andtape drives, but it can connect a wide range of other devices, including scanners andCDdrives, although not all controllers can handle all devices.
The ancestral SCSI standard, X3.131-1986, generally referred to as SCSI-1, was published by the X3T9 technical committee of theAmerican National Standards Institute (ANSI) in 1986. SCSI-2 was published in August 1990 as X3.T9.2/86-109, with further revisions in 1994 and subsequent adoption of a multitude of interfaces. Further refinements have resulted in improvements in performance and support for ever-increasing data storage capacity.[3]
SCSI is derived from theShugart Associates System Interface (SASI),[4] developed beginning 1979[5] and publicly disclosed in 1981.[1]Larry Boucher is considered to be the "father" of SASI and ultimately SCSI due to his pioneering work first at Shugart Associates and then atAdaptec, which he founded in 1981.[6]
A SASI controller provided a bridge between a hard disk drive's low-level interface and a host computer, which needed to read blocks of data. SASI controller boards were typically the size of a hard disk drive and were usually physically mounted to the drive's chassis. SASI, which was used in mini- and early microcomputers, defined the interface as using a 50-pin flat ribbon connector which was adopted as the SCSI-1 connector. SASI is a fully compliant subset of SCSI-1 so that many, if not all, of the then-existing SASI controllers were SCSI-1 compatible.[7]
In around 1980,NCR Corporation had been developing a competing interface standard by the name of BYSE. In the summer of 1981, NCR abandoned their in-house efforts in favor of pursuing SASI and improving on its design for their own computer systems. Fearing that their extension of the SASI standard would induce market confusion, however, NCR briefly cancelled their contract with Shugart. NCR's proposed improvements to the design of SCSI piqued the interest of Optimem, a subsidiary of Shugart, who requested that NCR and Shugart collaborate on a unified standard. In October 1981, the two companies agreed to co-develop SASI and present their standard jointly with ANSI.[8][9]
Until at least February 1982, ANSI developed the specification as "SASI" and "Shugart Associates System Interface".[10] However, the committee documenting the standard would not allow it to be named after a company. Almost a full day was devoted to agreeing to name the standard "Small Computer System Interface", which Boucher intended to be pronounced "sexy", but ENDL's[11] Dal Allan pronounced the new acronym as "scuzzy" and that stuck.[12]
The NCR facility inWichita, Kansas developed the industry's first SCSI controller chip, the NCR 5385, released in 1983. According to its developers, the chip worked the first time it was tested.[13][14] A number of companies, such as Adaptec and Optimem, were early supporters of SCSI.[10] By late 1990 at least 45 manufactures offered 251 models of parallel SCSI host adapters[15] Today, such host adapters have largely been displaced by the faster serial SCSI (SAS) host adapters.[16]
The "small" reference in "small computer system interface" is historical; since the mid-1990s, SCSI has been available on even the largest of computer systems.
Since its standardization in 1986, SCSI has been commonly used in theAmiga,Atari,Apple Macintosh andSun Microsystems computer lines and PC server systems. Apple started using the less-expensiveparallel ATA (PATA, also known asIDE) for its low-end machines with theMacintosh Quadra 630 in 1994, and added it to its high-end desktops starting with the Power Macintosh G3 in 1997. Apple dropped on-board SCSI completely in favor of IDE andFireWire with the (Blue & White) Power Mac G3 in 1999, while still offering aPCI SCSI host adapter as an option on up to the Power Macintosh G4 (AGP Graphics) models.[17] Sun switched its lower-end range toParallel ATA (PATA) with introduction of theirUltra 5 and 10 low end workstations usingCMD640 IDE controller and continued this trend with the laterBlade 100 and 150 entry level systems and did not switch to contemporarySATA interface even with the introduction of the Blade 1500 in 2003 while the higher end Blade 2500 released at the same time used Ultra320 Parallel SCSI-3. Sun moved toSATA andSAS interfaces with their last UltraSPARC-III based workstations in 2006 with the entry level Ultra 25 and mid-range Ultra 45. Commodore included SCSI on the Amiga 3000/3000T systems and it was an add-on to previous Amiga 500/2000 models. Starting with the Amiga 600/1200/4000 systems Commodore switched to the IDE interface. Atari included SCSI as standard in itsAtari MEGA STE,Atari TT andAtari Falcon computer models. SCSI has never been popular in the low-priced IBM PC world, owing to the lower cost and adequate performance of ATA hard disk standard. However, SCSI drives and even SCSIRAIDs became common in PC workstations for video or audio production.
Recent physical versions of SCSI—Serial Attached SCSI (SAS), SCSI-over-Fibre Channel Protocol (FCP), andUSB Attached SCSI (UAS)—break from the traditionalparallel SCSIbus and perform data transfer via serial communications usingpoint-to-point links. Although much of the SCSI documentation talks about the parallel interface, all modern development efforts use serial interfaces. Serial interfaces have a number of advantages over parallel SCSI, including higher data rates, simplified cabling, longer reach, improved fault isolation andfull-duplex capability. The primary reason for the shift to serial interfaces is theclock skew issue of high-speed parallel interfaces, which makes the faster variants of parallel SCSI susceptible to problems caused by cabling and termination.[18]
The non-physicaliSCSI preserves the basic SCSIparadigm, especially the command set, almost unchanged, through embedding of SCSI-3 overTCP/IP. Therefore, iSCSI useslogical connections instead of physical links and can run on top of any network supporting IP. The actual physical links are realized on lowernetwork layers, independently from iSCSI. Predominantly,Ethernet is used which is also of serial nature.
SCSI is popular on high-performance workstations, servers, and storage appliances. Almost all RAID subsystems on servers have used some kind of SCSI hard disk drives for decades (initially Parallel SCSI, interim Fibre Channel, recently SAS), though a number of manufacturers offerSATA-based RAID subsystems as a cheaper option. Moreover, SAS offers compatibility with SATA devices, creating a much broader range of options for RAID subsystems together with the existence ofnearline SAS (NL-SAS) drives. Instead of SCSI, modern desktop computers and notebooks typically use SATA interfaces for internal hard disk drives, withNVMe over PCIe gaining popularity as SATA can bottleneck modernsolid-state drives.
SCSI is available in a variety of interfaces. The first wasparallel SCSI (also called SCSI Parallel Interface or SPI), which uses aparallelbus design. Since 2005, SPI was gradually replaced bySerial Attached SCSI (SAS), which uses aserial design but retains other aspects of the technology. Many other interfaces which do not rely on complete SCSI standards still implement theSCSI command protocol; others drop physical implementation entirely while retaining theSCSI architectural model.iSCSI, for example, usesTCP/IP as a transport mechanism, which is most often transported overGigabit Ethernet or fasternetwork links.
SCSI interfaces have often been included on computers from various manufacturers for use underMicrosoft Windows,classic Mac OS,Unix,Amiga andLinux operating systems, either implemented on themotherboard or by the means of plug-in adaptors. With the advent ofSAS andSATA drives, provision for parallel SCSI on motherboards was discontinued.[19]
Initially, theSCSI Parallel Interface (SPI) was the only interface using the SCSI protocol. Its standardization started as asingle-ended 8-bitbus in 1986, transferring up to 5 MB/s, and evolved into a low-voltagedifferential 16-bit bus capable of up to 320 MB/s. The last SPI-5 standard from 2003 also defined a 640 MB/s speed which failed to be realized.
Parallel SCSI specifications include several synchronous transfer modes for the parallel cable, and an asynchronous mode. The asynchronous mode is a classic request/acknowledge protocol, which allows systems with a slow bus or simple systems to also use SCSI devices. Faster synchronous modes are used more frequently.
| Interface | Alternative names | Specification body / document | Width (bits) | Clock[a] | Line code | Maximum | ||
|---|---|---|---|---|---|---|---|---|
| Throughput | Length[b] | Devices[c] | ||||||
| Ultra-320 SCSI | Ultra-4; Fast-160 | SPI-5 (INCITS 367-2003) | 16 | 80 MHz DDR | none | 320 MB/s (2560 Mbit/s) | 12 m | 16 |
| SSA | Serial Storage Architecture | T10 / INCITS 309-1997 | serial | 200 Mbit/s | 8b10b | 20 MB/s[d][e][f] (160 Mbit/s) | 25 m | 96 |
| SSA 40 | T10 / INCITS 309-1997 | serial | 400 Mbit/s | 40 MB/s[d][e][f] (320 Mbit/s) | 25 m | 96 | ||
| Fibre Channel 1 Gbit | 1GFC | T11 / X3T11/94-175v0 FC-PH Draft, Revision 4.3 | serial | 1.0625 Gbit/s | 8b10b | 98.4 MB/s[e][f] (850 Mbit/s) | 500 m / 10 km[g] | 127 (FC-AL) 224 (FC-SW) |
| Fibre Channel 2 Gbit | 2GFC | T11 / X3T11/96-402v0 FC-PH-2, Rev 7.4 | serial | 2.125 Gbit/s | 197 MB/s[e][f] (1,700 Mbit/s) | 500 m / 10 km[g] | 127/224 | |
| Fibre Channel 4 Gbit | 4GFC | T11 / INCITS Project 2118-D / Rev 6.10 | serial | 4.25 Gbit/s | 394 MB/s[e][f] (3,400 Mbit/s) | 500 m / 10 km[g] | 127/224 | |
| Fibre Channel 8 Gbit | 8GFC | T11 / INCITS Project 2118-D / Rev 6.10 | serial | 8.5 Gbit/s | 788 MB/s[e][f] (6,800 Mbit/s) | 500 m / 10 km[g] | 127/224 | |
| Fibre Channel 16 Gbit | 16GFC | T11 / INCITS Project 2118-D / Rev 6.10 | serial | 14.025 Gbit/s | 64b66b | 1,575 MB/s[e][f] (13,600 Mbit/s) | 500 m / 10 km[g] | 127/224 |
| SAS 1.1 | Serial attached SCSI | T10 /INCITS 417-2006 | serial | 3 Gbit/s | 8b10b | 300 MB/s[e][f] (2,400 Mbit/s) | 6 m | 16,256[h] |
| SAS 2.1 | T10 /INCITS 478-2011 | serial | 6 Gbit/s | 600 MB/s[e][f] (4,800 Mbit/s) | 6 m | 16,256[h] | ||
| SAS 3.0 | T10 /INCITS 519 | serial | 12 Gbit/s | 1,200 MB/s[e][f] (9,600 Mbit/s) | 6 m | 16,256[h] | ||
| SAS 4.0 | T10 /INCITS 534 (draft) | serial | 22.5 Gbit/s | 128b150b | 2,400 MB/s[e][f] (19,200 Mbit/s) | tbd | 16,256[h] | |
| IEEE 1394-2008 | Firewire S3200, i.Link, Serial Bus Protocol (SBP) | IEEE Std. 1394-2008 | serial | 3.145728 Gbit/s | 8b10b | 315 MB/s (2,517 Mbit/s) | 4.5 m | 63 |
| SCSI Express | SCSI overPCIe (SOP) | T10 /INCITS 489 | serial | 8 GT/s (PCIe 3.0) | 128b130b | 985 MB/s[e][f][i] (7,877 Mbit/s) | short,backplane only | 258 |
| USB Attached SCSI 2 | UAS-2 | T10 /INCITS 520 | serial | 10 Gbit/s (USB 3.1) | 128b132b | ~1,200 MB/s[e][f] (~9,500 Mbit/s) | 3 m[j] | 127 |
| ATAPI overParallel ATA | ATA Packet Interface | T13 /NCITS 317-1998 at theWayback Machine (archived 2013-12-15) | 16 | 33 MHzDDR | none | 133 MB/s[k] (1,064 Mbit/s) | 457 mm (18 inches) | 2 |
| ATAPI overSerial ATA | serial | 6 Gbit/s | 8b10b | 600 MB/s[l] (4,800 Mbit/s) | 1 m | 1 (15 withport multiplier) | ||
| iSCSI | Internet Small Computer System Interface, SCSI overIP | IETF / RFC 7143 | mostly serial | implementation- and network-dependent | 1,187 MB/s[m] or 1,239 MB/s[n] | implementation- and network-dependent | 2128 (IPv6) | |
| SRP | SCSI RDMA Protocol (SCSI overInfiniBand and similar) | T10 /INCITS 365-2002 | implementation- and network-dependent | |||||
Internal parallel SCSI cables are usuallyribbons, with two or more 50–, 68–, or 80–pin connectors attached. External cables are typically shielded (but may not be), with 50– or 68–pin connectors at each end, depending upon the specific SCSI bus width supported. The 80–pinSingle Connector Attachment (SCA) is typically used for hot-pluggable devices
Fibre Channel can be used to transport SCSI information units, as defined by theFibre Channel Protocol for SCSI (FCP). These connections are hot-pluggable and are usually implemented with optical fiber.
Serial attached SCSI (SAS) uses amodifiedSerial ATA data and power cable.
iSCSI (Internet Small Computer System Interface) usually usesEthernet connectors and cables as its physical transport, but can run over any physical transport capable of transportingIP.
TheSCSI RDMA Protocol (SRP) is a protocol that specifies how to transport SCSI commands over a reliable RDMA connection. This protocol can run over any RDMA-capable physical transport, e.g.InfiniBand orEthernet when usingRoCE oriWARP.
USB Attached SCSI allows SCSI devices to use theUniversal Serial Bus.
The Automation/Drive Interface − Transport Protocol (ADT) is used to connect removable media devices, such as tape drives, with the controllers of the libraries (automation devices) in which they are installed. The ADI standard specifies the use ofRS-422 for the physical connections. The second-generation ADT-2 standard defines iADT, use of the ADT protocol over IP (Internet Protocol) connections, such as overEthernet. The Automation/Drive Interface − Commands standards (ADC, ADC-2, and ADC-3) define SCSI commands for these installations.
In addition to many different hardware implementations, the SCSI standards also include an extensive set of command definitions. The SCSI command architecture was originally defined forparallel SCSI buses but has been carried forward with minimal change for use with iSCSI and serial SCSI. Other technologies which use the SCSI command set include theATA Packet Interface,USB Mass Storage class andFireWire SBP-2.
In SCSI terminology, communication takes place between aninitiator and atarget. The initiator sends acommand to the target, which then responds. SCSI commands are sent in a Command Descriptor Block (CDB). The CDB consists of a one byte operation code followed by five or more bytes containing command-specific parameters.
At the end of the command sequence, the target returns a status code byte, such as 00h for success, 02h for an error (called a Check Condition), or 08h for busy. When the target returns a Check Condition in response to a command, the initiator usually then issues aSCSI Request Sense command in order to obtain a key code qualifier (KCQ) from the target. The Check Condition and Request Sense sequence involves a special SCSI protocol called a Contingent Allegiance Condition.
There are four categories of SCSI commands: N (non-data), W (writing data from initiator to target), R (reading data), and B (bidirectional). There are about 60 different SCSI commands in total, with the most commonly used being:
Each device on the SCSI bus is assigned a unique SCSI identification number or ID. Devices may encompass multiple logical units, which are addressed bylogical unit number (LUN). Simple devices have just one LUN, more complex devices may have multiple LUNs.
A "direct access" (i.e. disk type) storage device consists of a number of logical blocks, addressed by Logical Block Address (LBA). A typical LBA equates to 512 bytes of storage. The usage of LBAs has evolved over time and so four different command variants are provided for reading and writing data. The Read(6) and Write(6) commands contain a 21-bit LBA address. The Read(10), Read(12), Read Long, Write(10), Write(12), and Write Long commands all contain a 32-bit LBA address plus various other parameter options.
The capacity of a "sequential access" (i.e. tape-type) device is not specified because it depends, amongst other things, on the length of the tape, which is not identified in a machine-readable way. Read and write operations on a sequential access device begin at the current tape position, not at a specific LBA. The block size on sequential access devices can either be fixed or variable, depending on the specific device. Tape devices such as half-inch9-track tape,DDS (4 mm tapes physically similar toDAT),Exabyte, etc., support variable block sizes.
 | This sectionmay be too technical for most readers to understand. Pleasehelp improve it tomake it understandable to non-experts, without removing the technical details.(June 2008) (Learn how and when to remove this message) |
On a parallel SCSI bus, a device (e.g. host adapter, disk drive) is identified by a "SCSI ID", which is a number in the range 0–7 on a narrow bus and in the range 0–15 on a wide bus. On earlier models a physical jumper or switch controls theSCSI ID of the initiator (host adapter). On modern host adapters (since about 1997), doing I/O to the adapter sets the SCSI ID; for example, the adapter often contains anOption ROM (SCSI BIOS) program that runs when the computer boots up and that program has menus that let the operator choose the SCSI ID of the host adapter. Alternatively, the host adapter may come with software that must be installed on the host computer to configure the SCSI ID. The traditional SCSI ID for a host adapter is 7, as that ID has the highest priority during bus arbitration (even on a 16-bit bus).
The SCSI ID of a device in a drive enclosure that has a back plane is set either by jumpers or by the slot in the enclosure the device is installed into, depending on the model of the enclosure. In the latter case, each slot on the enclosure's back plane delivers control signals to the drive to select a unique SCSI ID. A SCSI enclosure without a back plane often has a switch for each drive to choose the drive's SCSI ID. The enclosure is packaged with connectors that must be plugged into the drive where the jumpers are typically located; the switch emulates the necessary jumpers. While there is no standard that makes this work, drive designers typically set up their jumper headers in a consistent format that matches the way that these switches implement.
Setting the bootable (or first) hard disk to SCSI ID 0 is an accepted IT community recommendation. SCSI ID 2 is usually set aside for the floppy disk drive while SCSI ID 3 is typically for a CD-ROM drive.[20]
Note that a SCSI target device (which can be called a "physical unit") is sometimes divided into smaller "logical units". For example, a high-end disk subsystem may be a single SCSI device but contain dozens of individual disk drives, each of which is a logical unit. Further, a RAID array may be a single SCSI device, but may contain many logical units, each of which is a "virtual" disk—a stripe set or mirror set constructed from portions of real disk drives. The SCSI ID, WWN, etc. in this case identifies the whole subsystem, and a second number, thelogical unit number (LUN) identifies a disk device (real or virtual) within the subsystem.
It is quite common, though incorrect, to refer to the logical unit itself as a "LUN".[21] Sometimes, redundantly, the actual LUN may be called a "LUN number" or "LUN id".[22]
In modern SCSI transport protocols, there is an automated process for the "discovery" of the IDs. The SSA initiator (normally the host computer through the 'host adaptor') "walk the loop" to determine what devices are connected and then assigns each one a 7-bit "hop-count" value.Fibre Channel – Arbitrated Loop (FC-AL) initiators use the LIP (Loop Initialization Protocol) to interrogate each device port for its WWN (World Wide Name). For iSCSI, because of the unlimited scope of the (IP) network, the process is quite complicated. These discovery processes occur at power-on/initialization time and also if the bus topology changes later, for example if an extra device is added.
SCSI has the CTL (Channel, Target or Physical Unit Number, Logical Unit Number) identification mechanism perhost bus adapter, or the HCTL (HBA, Channel, PUN, LUN) identification mechanism, one host adapter may have more than one channels.[23]
While all SCSI controllers can work with read/write storage devices, i.e. disk and tape, some will not work with some other device types; older controllers are likely to be more limited,[24] sometimes by their driver software, and more Device Types were added as SCSI evolved. Even CD-ROMs are not handled by all controllers. Device Type is a 5-bit field reported by aSCSI Inquiry Command; defined SCSI Peripheral Device Types include, in addition to many varieties of storage device, printer, scanner, communications device, and a catch-all "processor" type for devices not otherwise listed.
In larger SCSI servers, the disk-drive devices are housed in an intelligent enclosure that supportsSCSI Enclosure Services (SES). The initiator can communicate with the enclosure using a specialized set of SCSI commands to access power, cooling, and other non-data characteristics.
In 1979 the Company developed the first controller using the SASI interface ... in conjunction with Shugart Associates
NCR Corp., Microelectronics Division has announced the release of its NCR 5385 single-chip controller. NCR is the first to use the Small Computer System Interface (SCSI) protocol.
The lun command is used to create and manage luns[...]
If a LUN ID is not specified, the smallest number [...] is automatically picked.
