Article

Hard Drive Technologies

The concept of a hard drive as a large receptacle for computer data is understood by many. However, its inner workings and the methods by which hard drives are connected to computers is a bit more fuzzy. This article is a quick and dirty explanation of how hard drives work and the various types of methods by which they are attached to computers.

How do hard drives work?

A hard drive contains several round, flat disks called platters. These platters are stiff and are often called "disks". These stiff discs give hard drives their name (in the days of old, some discs were made of thin plastic; they were called Floppy Disks.) Platters, which normally measure 3.5-inches in size, are coated with a substance that allows data to be written magnetically. Mounted on a spindle, these platters are stacked on top of one another, separated by a cushion of air a scant millimeters apart. Within this space, the drive's read/write head passes over each of the platters and reads and writes data on each platter using electromagnetic signals. Heads can read and write on either side of each individual platter, depending on the drive's capacity; the more platters in a drive, the larger the capacity can be. Data is written in concentric circles (called tracks) that contain a number of sectors. Each of the sectors contains hundreds or thousands of bytes of data.

The platters sit on a spindle that is attached to a direct drive motor where the motor's shaft serves as the spindle. The spindle spins the platters thousands of rotations per minute, depending on the speed of the hard drive. Most consumer-based hard drives have spindle speeds of either 5,400RPM or 7,200RPM while faster server drives support speeds of 10,000RPM and 15,000RPM. This spindle speed is considered the mechanical speed of the drive; generally, the higher the speed, the faster the data can be accessed and read.

Areal Density

Spindle speed is only half the equation when determining hard drive speeds. Read speeds are also determined by areal density, or the amount of data that is packed into a square inch of platter space. The higher the areal density, the tighter the bits of data are packed, and the faster the read/write heads can read data. The measurement for areal density is gigabits per square inch (Gb/square inch) – keep in mind that 8 bits = 1 byte, so a hard drive with an areal density of 16 Gb/square inch has 2 gigabytes of data packed into every square inch of its platter. Add up all those numbers and then multiply by the number of platters (including both sides of the platter on some drives) and you can determine the total size of the drive. By the way, hard drives with areal densities of 16 Gb/square inch appeared in 2000; today, hard drives have areal densities approaching 100 Gb/square inch! So, a single 3.5" platter with an areal density of 100Gb/square inch could hold approximately 44GB of data on each side. This means that if both sides of the platter were used and a hard drive housed four of these platters, the overall capacity of that drive would be 350GB.

Internal Cache Buffer

The internal transfer rate of a hard drive is different than the external transfer rate. The internal rate is the rate at which the read/write heads can read data from the platters and then pass that data to the drive's onboard cache memory (also called a read buffer.) This cache memory is much faster than the mechanical drive. Cache memory stores data that the drive has previously read, so when you request data, your system checks the cache first before checking the actual drive which is much, much slower by comparison. Hard drive manufacturers have actually increased hard drive speeds by increasing the size of the internal cache buffers. With a larger 8 MB cache buffer, the computer system has a better chance of retrieving cached data instead of waiting for the read/write drive head to locate the data on the platter and then transfer it. The external transfer rate is the rate at which the drive can send the data from the onboard cache, across the drive's bus interface, through cabling, and into the system's main memory.

Various connection technologies (also called "buses") exist for hard drives and these will be discussed next. Keep in mind that while hard drive transfer rates are nowhere near the rates specified by these connection technologies. The industry tries to stay ahead of hard drive top-end speeds for various reasons. Currently, most new hard drives can only transfer 45-55 megabytes per second (MB/sec). However, hard drives can "burst" large amounts of data across the bus if the onboard cache has data the computer needs. These bursts can easily reach upwards of 120 MB/sec, so a good connection technology is essential for peak performance from modern hard drives.

IDE Hard Drive Connection Technology

IDE (Integrated Drive Electronics) refers to the controller on the drive itself – the brain behind the hard drive containing all information required to transfer data to the computer. Simply stated, without IDE, hard drives would be hard pressed to pass data back and forth with the computer – it would be lacking in intelligence. In the early days, before the days when IDE technology was located on the drive itself, hard drives required a separate host adapter to control the flow of data to the system.

But just what does a hard drive plug into and by what means is this connection accomplished? Hard drives are connected to the motherboard via EIDE sockets, which are in turn attached to an IDE controller chip on the motherboard. Today, two Enhanced IDE (EIDE) sockets are built onto most motherboards, supplying computers with two connections by which four hard drives can be used (two hard drives per socket.)

A 40-pin or 80-pin ribbon cable, depending on the speed of the hard drive (more on this later), is used to connect the drive to the system's motherboard. A power cable, typically denoted by four red, black and yellow wires with a white connector, is used to power the hard drive. This power connector is then connected to a special power supply that provides electricity to the computer.

What is ATA Technology?

Although the abbreviation IDE is popularly used in reference to hard drive interface connections, ATA (AT Attachment – AT refers to the Advanced Technology motherboard form factor) is also used. IDE and ATA are the same thing. The term Integrated Drive Electronics (IDE) is owned by Western Digital, so other companies, including Maxtor, Quantum and Seagate, use the term ATA instead. Upon reflection, it is interesting that most computer users refer to hard drives as IDE while manufacturers use the ATA term. Each of the ATA specifications is rated for different speeds and all ATA specifications are backwards compatible, so even the oldest ATA-1 hard drives can be run on the latest Ultra ATA buses.

Direct Memory Access

The term Ultra DMA also came to be known. DMA stands for Direct Memory Access and is a method, starting with ATA-2, by which data could be transferred directly to the computer's memory without the help of the processor, or without as much help. This method is also referred to as "Bus Mastering." DMA essentially frees up the processor to do other more important jobs like process data for applications. ATA-1, by comparison, used an older transfer method called Programmed Input/Ouput (PIO) that relies heavily on the CPU to transfer data to and from the hard drive. Hard drives still support PIO modes and revert to this older standard when the computer has driver problems or is running in safe mode. The difference in hard drive performance between PIO and DMA is noticeable.

The ATA Speed Limit

Since copper wire and copper interconnects are used to transfer electrical signals and data between chips and devices, inherent issues almost always crop up when data speeds and voltages are increased. With Ultra ATA/133, the limits of the original ATA specification have pretty much been reached. Crosstalk and signal noise, two interference issues that board designers battle even today when two copper interconnects are placed too close together. To fight this problem, starting with Ultra ATA/66 drives, 80-wire cables replaced the standard 40-wire ribbon cable used to connect hard drives to EIDE sockets. 80-wire cables contain 40 more ground wires that sit in-between the hot wires that carry data; these extra ground wires add insulation and stability to high-speed communications. Since the newer 80-wire cables have the same 40-hold plug, they fit into the same 40-pin socket. Detection technology is built into the Ultra ATA controller to detect which cable is used and which speeds are safe.

The ATA specification has other issues as well. Since motherboards only contain two EIDE sockets supporting two drives apiece, IDE is limited to two drives a chain. This is a problem since the ATA specification uses a parallel bus interface. Parallel buses are shared channels that transmit data over several wires simultaneously. As such, no other data can be transferred from other devices until the data from the first drive is finished. This can be an issue when two drives are connected to the same cable; they have to take turns to communicate. Also, IDE drives attached to the same cable must be configured as either master and slave. Pins on the drive are jumpered with little plastic connectors to specify whether the drive is the first or the second or slave drive so the computer knows which is which. This can be confusing to those wanting to add another hard drive to their system.

Serial ATA Hard Drive Connection Technology

Serial ATA, the latest hard drive connector technology, is an evolutionary replacement for the Parallel ATA interface. Serial ATA was developed and is administered and by The Serial ATA Working Group, an industry organization whose mission is to define, develop and deliver the industry specification for the Serial ATA interface. Unlike USB 2.0 and FireWire connectors, Serial ATA is planned to be the primary storage interface inside the PC system.

The first-generation of the Serial ATA standard supports a top-end speed of 150 MB/sec – which is not much faster than Ultra ATA/133. However, it was extremely unlikely that the Ultra-ATA 133 specification could make the jump to 266 megabytes per second. This is not a concern with Serial ATA. In fact, the second generation Serial ATA standard has already been ratified. Expect to see Serial ATA II supporting speeds of 300 megabytes per second and then doubling again to 600 megabytes per second with Serial ATA III. Speeds this high may seem like overkill, but they are required to keep pace with hard drive technologies. High-end workstations and servers take advantage of RAID technology. RAID (Redundant Array of Inexpensive/Independent Disks) is a method whereby hard drives can be connected such that data is interleaved in such a way that the overall performance improvement increase by a factor of the number of drives used. So, if you have three hard drives with transfer rates of 40 MB/sec apiece, the total combined RAID potential when combined is 120 MB/sec.

Serial ATA is completely compatible with today's software while providing many architectural enhancements. Using thinner wires that connect to only 4 pins rather than the 80 pins used today, and the lower voltages of Serial ATA cables allow them to be up to one meter in length compared to the 18" limitation of the 40- and 80-pin flat cable – almost three times the length. This means that Serial ATA wires are easier to route inside a computer and provide better airflow around parts. Also, the physical architecture change from parallel to serial means the master/slave designation is gone; the point-to-point connection of the Serial ATA interface requires each drive to be connected directly to the motherboard socket, so the days of daisy-chaining two hard drives off a single connection gone. Of course, since Serial ATA connectors are physically smaller than Parallel ATA sockets, more real estate is freed on motherboards and controller cards.

While Serial ATA drives are slow to surface, more and more motherboard manufacturers are including the new fangled connector in their designs. But don't worry about losing the older Parallel ATA connectors anytime soon as they will still be around for quite sometime before being phased out. Competitive desktop pricing will hopefully make Serial ATA hard drives and its architecture a widespread reality in the meantime.

SCSI Hard Drive Connection Technology

SCSI, pronounced “Scuzzy”, stands for Small Computer System Interface. It is a high-speed parallel interface standard that grew from a 20-page specification in 1980 into a complex, 600-page ANSI (American National Standards Institute) standard. The SCSI Trade Association was formed in 1996 to promote the use and understanding SCSI as a complete expansion bus that can be used to connect a wide variety of peripherals to a computer including hard drives, optical CD/DVD drives, tape backup drives internally and externally. SCSI peripherals require a dedicated SCSI host adapter, whether it be built directly into the motherboard or purchased separately as an expansion card. Due to the additional cost, SCSI is typically only built into workstation and server-class motherboards.

A lot of confusion exists between ATA and SCSI interfaces. ATA's biggest shortcoming is its limited cable length; at only 18 inches long, the standard is unsuitable for external devices. And while Serial ATA supports cable lengths up to 1 meter, it will probably never support external devices; this is especially true since USB 2.0 and FireWire are more than adequate for the job. But in the early days, before USB and FireWire, the only way to attach a high-performance hard drive externally was by using a SCSI connection.

Is SCSI better than ATA?

SCSI has many more strengths and features that make it compelling. For starters, SCSI supports Command Tag Queuing that allows SCSI devices to accept new commands while data that had already been read is sent back up the bus. This is in stark contrast to IDE’s single-threaded I/O interface that can only execute one command at a time and can only send data in one direction at a time. Command Tag Queuing allows the host adapter to execute up to 256 concurrent commands and in a different order than they were received. Serial ATA, on the other hand, will support a limited form of Command Queuing in its second-generation specification. Seagate built a limited form of it in their Barracuda line of Serial ATA hard drives, but it only supports up to 32 outstanding commands.

SCSI-based hard drives have faster spindle speeds and larger caches as they are designed for server applications. Video professionals covet SCSI drives for their high throughput, low CPU overhead, and smooth transfer rates. Also, SCSI can handles many more devices – in fact, modern SCSI host adapters can control up to 30 devices! Compare that to Ultra ATA’s 2 devices per channel (most motherboards only come with two) and it becomes apparent just how powerful SCSI is. SCSI is naturally better at performing data hungry tasks such as multi-tasking; this is because the SCSI bus controller is capable of controlling the drives without any work by the processor. Also, all drives on a SCSI chain are cable of operating at the same time. This was a very big performance advantage compared to Parallel ATA, but with the advent of Serial ATA, this advantage essentially disappears.

However, these features come at a price; SCSI costs more, especially after applying costs for the extra SCSI Host Adapter. SCSI is also a bit more difficult to install – the secret of SCSI is complete mastery of the elusive art of termination. Without proper termination, SCSI just doesn’t work. Termination works through resistors placed on either end of the SCSI chain, which consists of all the SCSI hard drives and devices connected to a SCSI host adapter. Older SCSI devices required a special resistor to be attached. Modern SCSI devices like hard drives, backup devices and optical drives have little plastic jumpers or switches to enable termination. SCSI cables are available with terminator blocks on the end of the cable as well. Also, every SCSI device must have a unique numbered ID which is set on each device by either setting jumpers or by using a built-in push-button dial. If the same ID is given to two devices or a devices receives the same ID as the host adapter, none of the SCSI devices will work. When SCSI doesn’t work, it is often due to poor termination or conflicting IDs.

The Flavors of SCSI

SCSI has continuously been upgraded many, many times; in fact, it is quite mind-boggling when one considers all of the different specifications. They include SCSI-1, SCSI-2, SCSI-3, Fast SCSI, Fast Wide SCSI, Ultra SCSI, Ultra Wide SCSI, Ultra2, Ultra160 and Ultra320 and iSCSI.

Ultra 3 SCSI ultimately became Ultra160 SCSI. This is because Ultra ATA/66 was becoming fashionable and the czars behind the SCSI standard began to learn the importance of marketing. So instead of referring SCSI by its specification number, it was decided to link SCSI with its connection speed while keeping the “Ultra” denotation. Hence, Ultra160 SCSI was born. And yes, it supports 160 MB/sec transfer speeds. Ultra160 does this by using a feature called double-edge clocking that doubles the transfer rate by taking advantage of both the rising and falling edges of the REQuest/ACKnowledge clock. This same technique has been popularized in Double Data Rate (DDR) memory. Since the Ultra160 accomplishes a bandwidth doubling without additional internal clocking, it reduces incompatibilities with older SCSI devices. Ultra160 SCSI also includes CRC (cyclical redundancy checking) for data integrity and uses the same LVD cabling as Ultra2 SCSI.

The latest flavor of SCSI is Ultra320 and it is the seventh-generation to be delivered on SCSI’s 20th birthday. Still only 16-bits wide, Ultra320 SCSI continues to adhere to backwards compatibility while evolving the SCSI standard at the same time. New features include packetized SCSI, Quick Arbitration and Selection (QAS), read and write data streaming and flow control, all of which decreases command overhead while further increasing performance. The last new standard is iSCSI, which is short for Internet SCSI. iSCSI is basically a method for transporting low latency SCSI blocks across IP networks. This makes iSCSI an ideal storage connection for storage area networks and for connecting large stores of data over the Internet because it provides to very important features that regular SCSI does not; high availability in case a line goes down and high scalability for the expansion of storage capacity without shutting down applications.