< Day Day Up > |
SCSI is the dominant I/O technology in the open systems world. Because it is difficult for new storage technologies to be integrated into the wide variety of products sold, new I/O technologies, such as Fibre Channel, are often based on some form of SCSI to facilitate rapid development. SCSI is a set of evolving ANSI-standard electronic interfaces that allow computers to communicate with peripheral hardware devices over a parallel bus. Although there have been recent changes to include serial connections, SCSI is still considered to be primarily a parallel bus. A diagram of the SCSI configuration architecture is shown in Figure 6-1. Figure 6-1. SCSI block diagram.Note For more information on SCSI technology, refer to The Book of SCSI, Second Edition: I/O for the New Millennium, by Gary Field and Peter M. Ridge; and Chapter 1, "The Network Storage Landscape," in Building Storage Networks, by Mark Farley. SCSI has evolved through three major standards beginning in 1986:
6.2.1 SCSI-1The original SCSI standard defined the first SCSI bus in terms of cable length, signal characteristics, commands, and transfer modes. The default speed for SCSI-1 was 5MB/s. It had an 8-bit parallel bus that transferred a single byte of data with each bus cycle. The features included support for two devices, asynchronous mode, and a single-ended interface. Regular and narrow conventions are no longer mentioned in the SCSI protocol names. 6.2.2 SCSI-2SCSI-2 was approved in 1994. SCSI-2 was an extensive enhancement that defined support for many advanced features, including the following:
SCSI-2 also maintained backward compatibility with all SCSI devices. 6.2.3 SCSI-3SCSI-3 was approved in 1996. SCSI-3 is a group of specifications that define the implementation of SCSI protocols on the following physical layers:
Each physical layer has different performance characteristics and uses different hardware. Other documents in the SCSI-3 standard continue to be developed. Currently, the SCSI-3 standard includes the SCSI-2 performance and functionality enhancements plus the following:
Note HP has extensively tested and integrated the Wide-Ultra SCSI-3 technology into HP servers and storage options because this technology allows the highest available performance in a SCSI host interface. In addition, backward compatibility provides investment protection for HP customers. 6.2.3.1 WIDE-ULTRA2 SCSI-3As the power of applications, processors, and storage devices increases, users seek ways to increase system performance. Wide-Ultra2 SCSI-3 products provide the speed, flexibility, and compatibility required to optimize workstations and servers. Figure 6-2 shows how the evolution of the SCSI standards have improved data transfer rates over the years. Figure 6-2. Data rate transfer speedsThe Wide-Ultra2 SCSI-3 card uses low-voltage differential (LVD) technology and features the latest SCSI specifications. The Wide-Ultra2 SCSI-3 protocol
6.2.3.2 WIDE-ULTRA2 SCSI-3 CONNECTORConventional attachment protocols that use separate power and data cables do not work well in high-end server environments. In addition to the more common cabling systems, high-end drives, usually Wide-Ultra or Ultra2 SCSI, can be set up to use a special connection system that replaces the power cable, data cable, and normal device ID assignment methods. Single Connector Attachment (SCA), which was developed for servers and other high-end systems, replaces the conventional cabling scheme with a backplane system that uses a single 80-pin connector, as shown in Figure 6-3. Figure 6-3. SCA-2 80-pin connector.This connector replaces the standard 68-pin Wide SCSI cable and the standard 4-wire D-shaped power connector. It also includes signals for setting the device IDs for individual drives. Single Connector Attachment-2 (SCA-2) connectors are an improved version of the original SCA connector. The major difference is that the SCA-2 connector is designed so that 2 of the 80 pins on the connector are always connected first when attaching a drive and disconnected last when removing a drive. This improves reliability by ensuring that the drive remains grounded. Multiple hard drives can be used together on high-end workstations and servers to increase performance and improve reliability. This typically is done using RAID, and in many cases, the hard drive subsystem is designed to allow hot swapping, which means that failed hard drives can be replaced without powering down the system. SCAs are hot-swappable. 6.2.3.3 ULTRA3The strategy for universal storage devices that span all ranges of storage systems is based on the industry-leading Wide-Ultra3 SCSI-3 technology. The Wide-Ultra3 SCSI-3 drives and options provide twice the speed of previous versions with data transfer rates up to 160MB/s. This SCSI specification is also known as Ultra160. The Ultra160 feature set is widely supported by HP and other industry-leading systems manufacturers as well as SCSI drive and component suppliers. As ratified by the SCSI Trade Association, the Ultra3 name refers to a product that incorporates any of the following features. 6.2.3.3.1 Double Transition ClockingBoth Wide-Ultra2 and Wide-Ultra3 use a 40MHz bus speed on a 16-bit wide bus. However, Wide-Ultra3 doubles the rate at which data is sampled during a given cycle. Therefore, Wide-Ultra3 has a maximum data transfer rate of 160MB/s. Figure 6-4 shows the Ultra2 data transfer with the normal clock cycle. Compare this with Figure 6-5, which shows the Ultra3 data transfer cycle using double-transition clocking. Figure 6-4. Ultra2 data transfer with normal clocking.Figure 6-5. Ultra3 data with double- transition clocking.6.2.3.3.2 Cyclical Redundancy Check (CRC)With faster transfer rates, the need for data reliability increases. CRC treats data as long binary numbers and divides the binary numbers by a prime number. Both the transmitter and receiver of the data perform the calculation. The remainder from the division is compared. In previous SCSI protocols, parity checking could detect only single-bit errors. CRC can detect single- and double-bit errors, an odd number of errors, and error bursts of up to 32 bits. 6.2.3.3.3 Domain ValidationThe host controller attempts to negotiate a data transfer rate until a successful connection is made to the target device or until all possible slower speeds have been attempted. During the negotiation process, the host controller sends certain data to a target device at a particular speed. If the target device returns identical data, the negotiated speed is used. Otherwise, the host controller attempts a reduced transfer rate until the connection is successful. With previous SCSI standards, devices that could not use a negotiated speed might have been inaccessible. For more information on Wide-Ultra3 technology, visit http://www.ultra160-scsi.com. For more information on HP ProLiant servers that support Wide-Ultra SCSI-3, visit http://www.compaq.com/smb/servers. For more information on Wide-Ultra3 SCSI, visit http://www.compaq.com/products/storageworks/ultra3/index.html. 6.2.3.4 ULTRA320Ultra320 SCSI, the next step in the SCSI evolution, introduces additional technologies, including protocol changes that reduce overhead and improve performance. These changes allow data to transfer safely and reliably at 320MB/s. Ultra320 SCSI includes the following key features:
Ultra320 SCSI maintains backward compatibility with previous versions of SCSI, even with the new added features. It also maintains compatibility with existing LVD SCSI technology, enabling customers to mix new and old technologies without interruption. For example, packetized SCSI is compatible with nonpacketized parallel SCSI. As a result, packetized SCSI devices can reside on the same bus as nonpacketized SCSI devices. The computer industry can continue to look forward to new and faster SCSI technology. Ultra640 is already in development. 6.2.3.4.1 Ultra320 SCSI Complements PCI-XThe faster Ultra320 SCSI I/O performance saturates the standard 66MHz PCI host bus that allowed for a maximum transfer rate of 533MB/s across a 64-bit bus, requiring PCI-X bus performance to take full advantage of Ultra320. Disk drive media data rates increase, exceeding 40MB/s. This required improvements over Ultra160 SCSI standard to support sustained throughput from the average number of drives in a server. With Ultra160 SCSI, two SCSI channels on a single device achieve a maximum transfer rate of 320MB/s, leaving sufficient overhead before saturating the PCI bus. However, at 320MB/s, two SCSI channels can achieve 640MB/s, which saturates a 64-bit/66MHz PCI bus. In addition to PCI-X doubling the performance of the host bus from 533MB/s to a maximum of 1066MB/s, there are protocol improvements so that efficiency of the bus is improved over PCI. Together PCI-X and Ultra320 SCSI provide the bandwidth necessary for current applications. 6.2.3.4.2 PrecompensationUltra320 SCSI speeds require new signaling technologies to maintain the high reliability required by server designs. Ultra320 SCSI signals on the SCSI bus are twice the frequency of Ultra160 SCSI signals, but cable requirements have not changed. Point-to-point connections can be 25 meters long, and multiple load systems can be 12 meters long. Doubling the maximum signal switching frequency in Ultra320 SCSI has pushed the SCSI bus into a frequency range that has greater signal attenuation in SCSI bus cables and has also required the signal slew rate to increase. The doubling of signal frequency has resulted in smaller amplitude signals and more reflections (undesired high frequency noise) on the SCSI bus. In addition to the attenuation problem, the effects of inter-symbol interference (ISI) in Ultra320 SCSI causes the bit edges of digital signals to be distorted from their true position. ISI is caused by long periods of the signal being at one voltage level, which charges the cable much like a capacitor. Skew compensation is required to address both the attenuation and ISI problems. In addition, Ultra320 SCSI requires precompensation (precomp) features in the SCSI output drivers to minimize attenuation and ISI. When precomp is enabled, the SCSI outputs switch to their maximum drive capability when there is a transition from 1 to 0 (or 0 to 1). The drive level is reduced on following bits when there is no transition. The reduction in drive level helps reduce the charging of the SCSI bus when the data bits contain a string of 0s or 1s, thereby reducing the ISI problems associated with an isolated 1 or an isolated 0. 6.2.3.4.3 Adjustable Active FilterAn option to use an adjustable active filter (AAF) to address the attenuation and ISI problem is included in Ultra320 SCSI. Precomp works to correct problems at the transmission end of the SCSI bus. AAF corrects problems on the receiving end of the bus. The purpose of adding an AAF to the SCSI receivers is to increase the signal-to-noise ratio of the SCSI signals. The AAF compensates for high-frequency attenuation in the cable and filters out the frequencies that are higher than the maximum Ultra320 SCSI signal frequency. The AAF automatically adjusts its high-frequency gain for the SCSI bus and calibrates itself during the training period so the high-frequency AAF gain (at 80MHz) cancels the high-frequency cable loss at 80MHz. Because Ultra320 SCSI has SCSI bus signals with switching frequencies up to 80MHz only, the AAF filters attenuates the unwanted frequencies (noise) more than 80MHz. Both precomp and AAF devices can reside on the same SCSI bus. However, when precomp is active, AAF is not required. If AAF is active, precomp is not required. The negotiation as to which technology is used occurs in the initialization process of the SCSI bus and is transparent to the user. 6.2.4 SCSI Standards SummaryThe following table summarizes the characteristics of each the 11 SCSI device types.
|
< Day Day Up > |