 |
Glossary
Index of all terms
Array
ASA
ASCII
Capacity
Compression
CPU
Data Compression
Data Transfer Rate
Density
Disc Storage
EISA
Fast Wide SCSI-2
Form Factor
Format
Gigabyte
Head
I/O
Input
Interface
Media
Megabyte
Operating System
Platter
Processing
RAID
RAW
Read
Resolution
SCSI
Sector
Silicon
Spindle
Spindle Motor
Step
Storage Capacity
Wide SCSI
Write
|
|
|
Barracuda 2 2HP: Parallel processing for storage devices |
|
|
| |
The conversion of raw data to a digital form has begun, and will only accelerate as new technologies emerge to take advantage of the malleability of digitized information. Voice text, graphics, video, audio, ASCII text and many other types of data are rapidly being converted to digital form. What is driving this trend is a convergence of technology and market demand for practicality in computing: computers, after all, are supposed to make information processing easier. The cost of technologies required to access "long" digital data forms is steadily decreasing, while users of information systems are increasing their dependence upon ever-increasing amounts of data. That data must be accessed, as well as transmitted, stored, manipulated and displayed ever more rapidly as user demands proliferate.
Despite the sudden, seemingly new demand for processing long, digitized data files, it is not a new form of data. Traditionally, digital data was only required in small blocks that tended to be repetitive, but this was more reflective of the applications in use and not the capabilities of a given system. Spreadsheets and word-processing documents are examples of such applications. A portion of a small 10- to 50-kilobyte file is called up, for example, displayed on a monitor and manipulated in digital form. All of the manipulation is done very slowly, however, when comparing a user's input to the speed of the actual system hardware.
Other factors contribute to the need for long data files that aren't related to applications but to the evolutionary makeup of the data. New digital data formats are much larger and sequential in nature: the best examples are audio and video files. Individual audio files can easily exceed a single Megabyte each, while an uncompressed, full-length feature movie takes nine Gigabytes or more of data to represent it. These long, sequential data types can easily tax a computing system's hardware capabilities.
Long files require computer systems to have ever greater processing power, faster I/O data rate performance and much more storage capacity than older, conventional data types. Fortunately, data compression techniques can be used to limit the amount of disc capacity needed for digital data. The drawback to compression when accessing and displaying large digital files is that they require complex processing techniques to uncompress and determine how the data is displayed or manifested to the user. The more the files are compressed, the more complex the calculations grow that are needed to process those compressed files (and the less representative or accurate the compressed data becomes). Meanwhile, for the system to practically process such large digital files the data must be transferred at relatively high speed from a storage device to the CPU. Finally, storage devices must have high enough capacity and performance to handle such large files to ensure the CPU can get enough data to maximize its own performance while processing these large digital files.
The processor
Today computer systems commonly employ one of two general processor architectures: CISC (Common Instruction Set Computer) or RISC (Reduced Instruction Set Computer). CISC processors have a large set of instructions in microcode (software programming that permanently resides within the chip) that are used in the process of completing a task. Most personal computers are built around Intel Corp.'s i386 and i486 microprocessors, which are CISC designs. RISC processors like Hewlett Packard's Precision Architecture RISC (PA-RISC) or IBM's RS/6000 have, for their part, been relegated to higher-performance computing applications like engineering workstations and networks that operate with the UNIX operating system.
Unlike CISC processors, RISC MPUs assume the majority of instructions that a CISC based processor must labor through to complete processing tasks are infrequently used. Therefore, RISC based processors have a smaller (reduced) set of instructions it follows to process similar tasks. By using a smaller instruction set, RISC-based designs are faster in their ability to process information than CISC designs. A 486 CISC processor can complete approximately 10 to 25 million instructions per second (MIPS). RISC designs that are targeted for use in workstations can easily complete up to 100 MIPS today.
A processor chain gang
The insatiable need to improve the raw performance of computing systems--which led to the small but growing adoption of RISC based designs--eventually led to another innovation for higher computing performance: the development of multiple-processor or multiprocessing systems. One type of multiprocessing is known as parallel processing, which involves breaking up complex instructions conventionally handled by a single processing and divvying up those instructions among multiple processors working at the same time. Overall system performance can be significantly increased with several independent processors, each with its own memory unit, and each working on individual elements of a single complex problem.
The other element
Magnetic hard disc drives are the mass storage device of choice for performance and capacity when dealing with large file computing applications. Other popular storage devices include optical disc drives (CD-ROM, rewriteable magneto optical, and write-once/read many or WORM), which can store very large files on relatively low cost removeable media, are not as fast as magnetic hard disc drives are. Most optical drives in the market today are unable to sustain the data rates necessary for most uncompressed audio and video applications, although data compression does alleviate some of that performance burden. Silicon options such as static-RAM and flash memory chips, which easily have the performance characteristics needed to process such files, are still prohibitively expensive for storing large files in mainstream applications. Hard disc storage has the best combination of performance, capacity and cost available to the market today for handling these types of files.
Despite the clear advantages of disc drive storage over other available storage technologies, long data formats can still create problems when the data rates required by an application exceed the capabilities of a drive. A disc drive is comprised of one or more platters that each have a read/write head per surface.
Data is stored on the platter surface and retrieved or stored by the read/write head as the platter rotates. The rate at which data is retrieved or stored, data rate, is generally fixed by two factors, the speed of the platter rotation and the amount of data that is packed onto an individual platter. The faster the platter spins and the denser the data is on the media, the better the performance potential for the drive. A spin rate of 7,200 RPM and density of 300 Megabits per square inch are common features of today's high-performance drives, which yield data rates of about 6 Megabytes per second.
Some solutions have been developed to meet the need for higher data rates such as RAID (redundant array of inexpensive discs) subsystems. Data compression, as mentioned above, is another common solution. Both solutions have drawbacks, however. RAID subsystems require multiple drives and host adapters, and the cost of that hardware adds up quickly. The use of compression lowers the resolution of an image and requires additional hardware performance support. Data rates in excess of 10 MB/sec were possible only through inventive, complicated and proprietary solutions, not with a standard-interface disc drive. That is, not until the development of dual-head parallel technology in 3.5-inch SCSI disc drives.
Dual-head parallel development
Dual-head parallel technology isn't a new development in disc drives. The use of multiple-head parallel technology was developed by Control Data Corp.'s Imprimis storage unit (which was acquired by Seagate Technology in 1990). Imprimis introduced a four-head parallel disc drive in 1987 it called the Hydra. Its next product was the first dual-head parallel drive to reach the market, the Sabre 5 2HP, an 8-inch drive which was introduced in 1989 and designed for supercomputing applications. The dual-head design concept has been field-proven in seven models since 1987 ranging from the 14-inch Hydra 4, to the leading-edge 3.5-inch Barracuda 2 2HP.
Parallel head technology is based on a concept similar to that of parallel processors. Large files are divided into smaller pieces which are distributed on separate disc drive platters. Unlike standard disc drives we examined earlier, which can only read or write from one head at a given point in time, parallel head drives can read or write from two or more heads at the same time. The following diagram details the process for a dual head parallel design with a Fast Wide SCSI-2 interface.
Several advancements in technology, primarily in the SCSI interface and read-channel circuits, have accelerated the evolution and migration of dual-head technology from high-end supercomputer storage devices to disc drives for personal computers.
Faster and smaller
The performance of SCSI itself has improved over the years as it evolved through successive generations. Early SCSI disc drives used "standard" SCSI, which was limited to a 5 Mbyte per second data transfer rate. The next significant performance step was the introduction of Fast SCSI-2, which boosted data transfer rates to 10 Mbytes per second. Most disc drives over a Gigabyte in capacity have Fast SCSI-2, enabling connection to the majority of modern computing platforms.
The performance requirements of new applications involving video data can now exceed the data transfer rate allowed by the popular Fast SCSI-2 interface. This dilemma has required the development of the Fast Wide SCSI-2 interface, which allows data transfers up to 20 Mbytes per second. Other extensions of SCSI, such as 40 Mbyte/second Ultra SCSI technology, are also in development, but the bandwidth and widespread availability of Fast Wide SCSI-2 has made it the interface of choice for use in dual-head parallel disc drives.
One technology advancement that Seagate has exclusively employed in its parallel head architecture is the use of a single SCSI integrated circuit that enables the separation or combination of data as it is sent to and from a disc and the host system. This SCSI chip has two serializers/deserializers which perform the task of data seperation or combination and two (?) FIFO (First-In, First-Out) buffers that queue up processed data for transfer to the disc or host.
These advancements in technology, Fast Wide SCSI-2 and innovative, powerful ICs, created the opportunity to transfer the parallel head technology to a common design platform. For Seagate, the high performance Barracuda disc drive family was the logical choice as the base platform for dual-head technology.
Barracuda
The Barracuda family of drives were the first production drives in the world to achieve the technically elusive goal of a 7,200 RPM spin rate. The first model, the 3.5-inch form factor Barracuda 2, ran at 7,200 RPM and had a storage capacity of 2.1 Gbytes. The high RPM was made possible by using flex circuits to conserve space and thereby allow the use of a taller spindle motor. The taller spindle motor, with more coil windings than a standard 3.5-inch disc drive motor, enabled higher RPMs. The tall spindle also enabled the design of a more power-efficient motor, reducing the power and wattage dissipation associated with faster spindle speeds.
Barracuda 2 2HP
Seagate then combined Fast Wide SCSI and the new SCSI ICs with the Barracuda 2 design, and that resulted in the Barracuda 2 2HP, a drive that realized a tremendous increase in data transfer rate performance. For example, and using a sector size of about 2 Mbytes for comparison, the formatted data rate of the Barracuda 2 is about 3.9- to 6.6 Mbytes/sec.; the data rate of the Barracuda 2 2HP ranges from 7.6 to 12.8 Mbytes/sec, a substantial increase in performance. The peak data transfer rate for the Barracuda 2 2HP is 14.1 Mbytes per second, double the throughput of a standard disc drive. The Barracuda 2 2HP's high data rate is ideal for use in long or large digital data applications, such as image processing for photographic and pre-press purposes, video-on-demand and other multimedia systems.
There are several important benefits to the Barracuda 2 2HP beyond the high data rate. It uses the industry-standard Fast Wide SCSI-2 and is based on a field-proven design. A large number of Fast Wide SCSI-2 host adapters have been developed to support a variety of system buses, such as S-bus, EISA, PCI and VL-bus applications. These adapters enable the quick integration of the Barracuda into many high-performance computing platforms.
The use of the Barracuda design leverages the knowledge gained through hundreds of thousands of hours of testing. Additionally, leading-edge technologies incorporated in the Barracuda, such as the previously-mentioned flex circuitry and tall spindle motor, have proven its reliability not only in the labs but in the field at thousands of user sites worldwide. Other technologies employed in the drive, such as the Seagate-developed Advanced SCSI Architecture (ASA), allows multiprocessing of SCSI commands for even faster performance and has designed into more than one million Seagate products. The Barracuda's solid and highly reliable design is backed by a five-year warranty.
The Barracuda 2 2HP is a leading-edge solution to leading-edge problems associated with processing large new data types. The use and need for long forms of data will only increase as new applications continue to become available and users become adept with them. The Barracuda 2 2HP, compared to RAID subsystems and other complex storage solutions, is the most effective solution to provide the necessary performance and storage capacity long file users require.
Digital future
The digital format, meanwhile, allows historically unprecedented access to data. The cost of transmitting, storing, manipulating and displaying digital format data is decreasing steadily. The continuous reduction in cost will open up new markets, exemplified by the proliferation of the multimedia PC into the office and now the home. Market requirements will demand computers to become ever easier to use, which means the hardware and software components of those systems will more technologically complex. The Barracuda 2 2 HP disc drive, with its complex, parallel-head design and easily integratable Fast Wide SCSI-2 interface will become the standard for long data processing requirements today and into the near future.
|
|
|
|