Anatomy of a Hard Drive

A hard drive is constructed in a cleanroom to avoid the introduction of contaminants into the hermetically sealed drive casing. Once the casing is sealed, most manufacturers seal one or more of the screws with a sticker warning that removal of or damage to the seal will result in voiding the drive’s warranty. Even some of the smallest contaminants can damage the precision components if allowed inside the hard drive’s external shell. The following is a list of the terms used to describe these components in the following paragraphs:

• Platters
• Read/write heads
• Tracks
• Sectors
• Cylinders
• Clusters (allocation units)

Inside the sealed case of the hard drive lie one or more platters, where the actual data is stored by the read/write heads. The heads are mounted on a mechanism that moves them in tandem across both surfaces of all platters. Older drives used a stepper motor to position the heads at discrete points along the surface of the platters, which spin at thousands of revolutions per minute on a spindle mounted to a hub. Newer drives use voice coils for a more analog movement, resulting in reduced data loss because the circuitry can sense where the data is located through a servo scheme, even if the data shifts due to changes in physical disc geometry. Figure 2.3 illustrates the key terms presented in this discussion. The four stacked discs shown in the illustration are platters.

Factory preparation for newer drives, or low-level formatting in the field for legacy drives, maps the inherent flaws of the platters so that the drive controllers know not to place data in these compromised locations. Additionally, this phase in drive preparation creates concentric rings, or tracks, which are drawn magnetically around the surface of the platters. Sectors are then delineated within each of the tracks. Sectors are the magnetic domains that represent the smallest units of storage on the disk’s platters. Magnetic-drive sectors commonly store only 512 bytes (1/2KB) of data each.

The capacity of a hard drive is a function of the number of sectors it contains. The controller for the hard drive knows exactly how the sectors are laid out within the disk assembly. It takes direction from the BIOS when writing information to and reading information from the drive. The BIOS, however, does not always understand the actual geometry of the drive. For example, the BIOS does not support more than 63 sectors per track. Nevertheless, many hard drives have tracks that contain many more than 63 sectors per track. As a result, a translation must occur from where the BIOS believes it is directing information to be written to where the information is actually written by the controller. When the BIOS detects the geometry of the drive, it is because the controller reports dimensions that the BIOS can understand. The same sort of trickery occurs when the BIOS reports to the operating system a linear address space for the operating system to use when requesting that data be written to or read from the drive through the BIOS.

The basic hard drive geometry consists of three components: the number of sectors that each track contains, the number of read/write heads in the disk assembly, and the number of cylinders in the assembly. This set of values is known as CHS (for cylinders/heads/sectors). The number of cylinders is the number of tracks that can be found on any single surface of any single platter. It is called a cylinder because the collection of all same-number tracks on all writable surfaces of the hard drive assembly looks like a geometric cylinder when connected together vertically. Therefore, cylinder 1, for instance, on an assembly that contains three platters comprises six tracks (one on each side of each platter), each labeled track 1 on its respective surface.

Because the number of cylinders indicates only the number of tracks on any one writable surface in the assembly, the number of writable surfaces must be factored into the equation to produce the total number of tracks in the entire assembly. This is where the number of heads comes in. There is a single head dedicated to each writable surface, two per platter. By multiplying the number of cylinders by the number of heads, you produce the total number of tracks throughout the disk assembly. By multiplying this product by the number of sectors per track, you discover the total number of sectors throughout the disk assembly. Dividing the result by 2 provides the number of kilobytes that the hard drive can store. This works because each sector holds 512 bytes, which is equivalent to 1/2KB. Each time you divide the result by 1024, you obtain a smaller number, but the unit of measure increases from kilobytes to megabytes, from megabytes to gigabytes, and so on. The equation in Figure 2.4 illustrates this computation:

For example, a drive labeled with the maximum allowed CHS geometry of 16383/16/63, respectively, results in only 7.9GB. Using the equation and multiplying the number of cylinders by the number of heads, you arrive at 262,128 total tracks on the drive. Multiplying this number by 63, the result is that there are 16,514,064 total sectors on the drive. Each sector holds 1/2KB, for a total capacity of 8,257,032KB. Dividing by 1024 to convert to MB and again by 1024 to convert to GB, the 7.9GB capacity is revealed. As a result, although drives larger than 8GB still display the 16383/16/63 CHS capacity for devices that must adhere to the CHS geometry, the CHS scheme cannot be used on today’s larger drives at the risk of losing the vast majority of their capacity. The solution is to allow the operating system to reference logical blocks of 1/2KB sectors that can be individually addressed by a 48-bit value, resulting in 128PB of drive capacity, far above the largest drives being manufactured today. A PB is 1024TB; a TB is 1024GB.

File systems laid down on the tracks and their sectors routinely group a configurable number of sectors into equal or larger sets called clusters or allocation units. This concept exists because operating system designers have to settle on a finite number of addressable units of storage and a fixed number of bits to address them uniquely. Because the units of storage can vary in size, however, the maximum amount of a drive’s storage capacity can vary accordingly, but not unless logical drive capacities in excess of 2TB are implemented. Volumes based on the master boot record (MBR) structure are limited to 2TB total. Such volumes include those created on basic disks as well as simple and mirrored volumes on dynamic disks. Be aware that today’s hard drives and volumes created with RAID can certainly exceed 2TB by combining multiple simple volumes into spanned or striped volumes, at which point maximum NTFS volume sizes—discussed at the end of this section—come into play. Keep in mind that larger clusters beget larger volumes, but they result in less efficient usage of space, a phenomenon that is discussed in the following paragraph.

No two files are allowed to occupy the same sector, so the opportunity exists for a waste of space that defragmenting cannot correct. Clusters exacerbate the problem by having a similar foible: The operating system does not allow any two files to occupy the same cluster. Thus the larger the cluster size, the larger the potential waste. So although you can increase the cluster size (generally to as large as 64KB, which corresponds to 128 sectors), you should keep in mind that unless you are storing a notable number of very large files, the waste will escalate astoundingly, perhaps negating or reversing your perceived storage-capacity increase. Nevertheless, if you have single files that are very large, say hundreds of megabytes or larger, then increased cluster sizes are for you.

For instance, assuming the 2TB maximum size for simple and mirrored dynamic volumes, a 64KB cluster size results in a maximum spanned or striped NTFS volume size of 64KB less than 256TB.

HDD Speeds

As the electronics within the HBA and controller get faster, they are capable of requesting data at higher and higher rates. If the platters are spinning at a constant rate, however, the information can only be accessed as fast as a given fixed rate. To make information available to the electronics more quickly, manufacturers increase the speed at which the platters spin from one generation of drives to the next, with multiple speeds coexisting in the marketplace for an unpredictable period, at least until the demand dies down for one or more speeds.

The following spin rates have been used in the industry for the platters in conventional magnetic hard disk drives:

• 5400 rpm
• 7200 rpm
• 10,000 rpm
• 12,000 rpm

While it is true that a higher revolutions per minute (rpm) rating results in the ability to move data more quickly, there are many applications that do not benefit from increased disk-access speeds. As a result, you should choose only faster drives, which are also usually more expensive per byte of capacity, when you have an application for this type of performance, such as for housing the partition where the operating system resides or for very disk-intensive programs. The lower speeds can be ideal in laptops, where heat production and battery usage can be issues with higher-speed drives.

Standard SSDs

In contrast, solid-state drives (SSDs) have no moving parts, but they use the same solid-state memory technology found in the other forms of flash memory. All solid-state memory is limited to a finite number of write (including erase) operations. Algorithms have been developed to spread the write operations over the entire device constantly. Such “wear leveling” increases the life of the SSD, but lack of longevity remains a disadvantage of this technology.

SSDs read contents more quickly, can consume less power and produce less heat, and are more reliable and less susceptible to damage from physical shock and heat production than their magnetic counterparts. However, the technology to build an SSD is still more expensive per byte, and SSDs are not yet available in capacities high enough to rival the upper limits of conventional hard disk drive technology. As of August 2015, for instance, the largest SSD on the market was 2TB, while the largest readily available conventional HDD was 8TB.

SSDs are separated into two broad categories: volatile DRAM-based and nonvolatile flash-based. Flash-based SSDs made with NAND memory (a transistor-based gate that has the opposite output to an AND gate) use considerably less power than HDDs. Those made with DRAM can use every bit as much power as conventional drives, however. The advantage of those made with the standard RAM modules used in desktop motherboards is that the modules can often be upgraded using larger modules, making a larger SSD overall.

When used as a replacement for traditional HDDs, SSDs are most often expected to behave in a similar fashion, mainly by retaining contents across a power cycle. With SSD, you can also expect to maintain or exceed the speed of the HDD. You can compensate for the volatility of DRAM-based SSDs by adding a backup power source, such as a battery or capacitor, or by keeping a nonvolatile backup of the drive’s data that does not detract from the speed of the primary storage location. Flash-based SSDs, while faster during read operations than their HDD counterparts, can be made faster still by including a small amount of DRAM as a cache. DRAM-based SSDs are faster still.

Hybrid Drives

A cost-saving alternative to a standard SSD that can still provide a significant increase in performance over conventional HDDs is the hybrid drive. Hybrid drives can be implemented in two ways: a solid-state hybrid drive and a dual-drive storage solution. Both forms of hybrid drives can take advantage of solutions such as Intel’s Smart Response Technology (SRT), which inform the drive system of the most used and highest-value data. The drive can then load a copy of such data into the SSD portion of the hybrid drive for faster read access.

It should be noted that systems on which data is accessed randomly do not benefit from hybrid drive technology. Any data that is accessed for the first time will also not be accessed from flash memory, and it will take as long to access it as if it were accessed from a traditional hard drive. Repeated use, however, will result in the monitoring software’s flagging of the data for caching in the SSD.

CD-ROMs, DVD-ROMs, BD-ROMs, and Capacities

The amount of data that can be stored on the three primary formats of optical disc varies greatly, with each generation of disc exceeding the capacity of all previous generations. The following sections detail the science behind the capacities of all three formats.

DVD-ROM

For even more storage capacity, many computers feature some form of DVD or BD drive, such as the original DVD-ROM drive. The basic DVD-ROM disc is a single-sided disc that has a single layer of encoded information. These discs have a capacity of 4.7GB, many times the highest CD-ROM capacity. Simple multiplication can sometimes be used to arrive at the capacities of other DVD-ROM varieties. For example, when another media surface is added on the side of the disc where the label is often applied, a double-sided disc is created. Such double-sided discs have a capacity of 9.4GB, exactly twice that of a single-sided disc.

Practically speaking, the expected 9.4GB capacity from two independent layers isn’t realized when those layers are placed on the same side of a DVD, resulting in only 8.5GB of usable space. (BDs do not have this issue; they make use of the full capacity of each layer.) The loss of capacity is due to the space between tracks on both layers being 10 percent wider than normal to facilitate burning one layer without affecting the other. This results in about 90 percent remaining capacity per layer. This technology is known as DVD DL (DL for dual-layer), attained by placing two media surfaces on the same side of the disc, one on top of the other, and using a more sophisticated mechanism that burns the inner layer without altering the semitransparent outer layer and vice versa, all from the same side of the disc. Add the DL technology to a double-sided disc, and you have a disc capable of holding 17.1GB of information—again twice the capacity of the single-sided version. Figure 2.5 shows an example of an early DVD-ROM drive, which also accepts CD-ROM discs. Modern 5¼″ optical drives are indistinguishable from older ones, aside from obvious markings concerning their capabilities.

Optical Drive Data Rates

CD-ROM drives are rated in terms of their data transfer speed. The first CD-ROM drives transferred data at the same speed as home audio CD players, 150KBps, referred to as 1X. Soon after, CD drives rated as 2X drives that would transfer data at 300KBps appeared. They increased the spin speed in order to increase the data transfer rate. This system of ratings continued up until the 8X speed was reached. At that point, the CDs were spinning so fast that there was a danger of them flying apart inside the drive. So, although future CD drives used the same rating (as in 16X, 32X, and so on), their rating was expressed in terms of theoretical maximum transfer rate; 52X is widely regarded as the highest multiplier for data CDs. Therefore, the drive isn’t necessarily spinning faster, but through electronics and buffering advances, the transfer rates continued to increase.

The standard DVD-ROM 1X transfer rate is 1.4MBps, already nine times that of the comparably labeled CD-ROM. As a result, to surpass the transfer rate of a 52X CD-ROM drive, a DVD-ROM drive need only be rated 6X. DVD transfer rates of 16X at the upper end of the scale are common.

The 1X transfer rate for Blu-ray is 4.5MBps, roughly 3¼ times that of the comparable DVD multiplier and close to 30 times that of the 1X CD transfer rate. It takes 2X speeds to play commercial Blu-ray videos properly.

Recordable Discs and Burners

Years after the original factory-made CD-ROM discs and the drives that could read them were developed, the industry, strongly persuaded by consumer demand, developed discs that, through the use of associated drives, could be written to once and then used in the same fashion as the original CD-ROM discs. The firmware with which the drives were equipped could vary the power of the laser to achieve the desired result. At standard power, the laser allowed discs inserted in these drives to be read. Increasing the power of the laser allowed the crystalline media surface to be melted and changed in such a way that light would reflect or refract from the surface in microscopic increments. This characteristic enabled mimicking the way in which the original CD-ROM discs stored data.

Eventually, discs that could be written to, erased, and rewritten were developed. Drives that contained the firmware to recognize these discs and control the laser varied the laser’s power in three levels. The original two levels closely matched those of the writable discs and drives. The third level, somewhere in between, could neutralize the crystalline material without writing new information to the disc. This medium level of power left the disc surface in a state similar to its original, unwritten state. Subsequent high-power laser usage could write new information to the neutralized locations.

The best algorithms for such drives, which are still available today, allow two types of disc erasure. The entire disc can be erased before new data is written (erased or formatted, in various application interfaces), or the data can be erased on the fly by one laser, just fractions of a second before new data is written to the same location by a second laser. If not properly implemented in a slow, determined fashion, the latter method can result in write errors because the crystalline material does not adequately return to its neutral state before the write operation. The downside to slowing down the process is obvious, and methods exist to allow a small level of encoding error without data loss. This need to move more slowly adds a third speed rating, the rewrite speed, to the existing read and write speed ratings of a drive. The following section delves more deeply into this concept. Updates to the drive’s firmware can often increase or equalize these speeds.

Recordable CD Formats

CD-recordable (CD-R) and CD-rewritable (CD-RW) drives (also known as CD burners) are essentially CD-ROM drives that allow users to create (or burn) their own CD-ROMs. They look very similar to CD-ROM drives, but they feature a logo on the front panel that represents the drive’s CD-R or CD-RW capability. Figure 2.6 shows the CD-R and CD-RW logos that you are likely to see on such drives.

The difference between these two types of drives is that CD-R drives can write to a CD-R disc only once. A CD-RW drive can erase information from a CD-RW disc and rewrite to it multiple times. Also, CD-RW drives are rated according to their write, rewrite, and read times. Thus, instead of a single rating like 64X, as in the case of CD-ROM drives, CD-RW drives have a compound rating, such as 52X-32X-52X, which means that it writes at 52X, rewrites at 32X, and reads at 52X.

Recordable DVD Formats

A DVD burner is similar to a CD-R or CD-RW drive in how it operates: It can store large amounts of data on a special, writable DVD. Single-sided, dual-layer (DL) discs can be used to write 8.5GB of information to one single-sided disc. Common names for the variations of DVD burning technologies include DVD+R, DVD+RW, DVD-R, DVD-RW, DVD-RAM, DVD-R DL, and DVD+R DL. The “plus” standards come from the DVD+RW Alliance, while the “dash” counterparts are specifications of the DVD Forum. The number of sectors per disc varies between the “plus” and “dash” variants, so older drives might not support both types. The firmware in today’s drives knows to check for all possible variations in encoding and capability. The “plus” variants have a better chance of interoperability, even without the disc being finalized.

A DVD-ROM or recordable drive looks very similar to a CD-ROM drive. The main difference is the presence of one of the various DVD logos on the front of the drive. CD-ROM and recordable CDs can usually be read and, if applicable, burned in DVD burners. Figure 2.7 and Figure 2.8 show the most popular data-oriented logos that you are likely to see when dealing with DVD drives suited for computers. Figure 2.7 shows the “dash” logos, while Figure 2.8 shows the “plus” logos.

Table 2.2 lists the main DVD formats used for storing and accessing data in computer systems as well as their characteristics.

Format	Characteristics
DVD-ROM	Purchased with data encoded; unable to be changed.
DVD-R, DVD+R	Purchased blank; able to be written to once and then treated like a DVD-ROM.
DVD-RW, DVD+RW	Purchased blank; able to be written to and erased multiple times; session usually must be closed for subsequent access to stored data.

Recordable BD Formats

The Blu-ray Disc Association duplicated the use of the R suffix to denote a disc capable of being recorded only once by the consumer. Instead of the familiar RW, however, the association settled on RE, short for re-recordable. As a result, watch for discs labeled BD-R and BD-RE. Dual-layer versions of these discs can be found as well.

The Blu-ray Disc Association decided against creating separate logos for each BD type, resolving instead to use only the logo shown in Figure 2.9. Discs are labeled most often in a sans-serif font with the actual type of the disc as well as this generic BD logo.

Serial AT Attachment Drives

At one time, integrated drive electronics (IDE) drives were the most common type of hard drive found in computers. Though often thought of in relation to hard drives, IDE was much more than a hard drive interface; it was also a popular interface for many other drive types, including optical drives and tape drives. Today, we call it IDE PATA and consider it to be a legacy technology. The industry now favors SATA instead.

Serial ATA began as an enhancement to the original ATA specifications, also known as IDE and, today, PATA. Technology is proving that the orderly progression of data in a single-file path is superior to placing multiple bits of data in parallel and trying to synchronize their transmission to the point that each bit arrives simultaneously. In other words, if you can build faster transceivers, serial transmissions are simpler to adapt to the faster rates than are parallel transmissions.

The first version of SATA, known as SATA 1.5Gbps (and also by the less-preferred terms SATA I and SATA 150), used an 8b/10b-encoding scheme that requires 2 non-data overhead bits for every 8 data bits. The result is a loss of 20 percent of the rated bandwidth. The silver lining, however, is that the math becomes quite easy. Normally, you have to divide by 8 to convert bits to bytes. With 8b/10b encoding, you divide by 10. Therefore, the 150MBps throughput for which this version of SATA was nicknamed is easily derived as 1/10 of the 1.5Gbps transfer rate. The original SATA specification also provided for hot swapping at the discretion of the motherboard and drive manufacturers.

Similar math works for SATA 3Gbps, also recklessly tagged as SATA II and SATA 300, and SATA 6Gbps, which is not approved for being called SATA III or SATA 600, but the damage is already done. Note that each subsequent version doubles the throughput of the previous version. Figure 2.10 shows a SATA connector on a data cable followed by the headers on a motherboard that will receive it. Note that identifiers silkscreened onto motherboards often enumerate such headers. The resulting numbers are not related to the SATA version that the header supports. Instead, such numbers serve to differentiate headers from one another and to map to firmware identifiers, often visible within the BIOS configuration utility.

The card-edge style connectors for data and power are arranged in such a manner on the back of SATA drives that no cables are required, although desktop and server systems almost certainly employ them. The same interface, however, can be used in laptops without the adapters needed to protect the delicate pins of the parallel interfaces found on the preceding generation of small form factor drives. The lack of adapter also leads to less space reserved in the bays for drives of the same size, giving designers and consumers the choice between smaller systems or more complex circuitry that can move into the newly available space.

RAID

RAID stands for Redundant Array of Independent Disks. It’s a way of combining the storage power of more than one hard disk for a special purpose, such as increased performance or fault tolerance. RAID can be implemented in software or in hardware, but hardware RAID is more efficient and offers higher performance but at an increased cost.

There are several types of RAID. The following are the most commonly used RAID levels:

RAID 0 Also known as disk striping, where a striped set of equal space from at least two drives creates a larger volume. This is in contrast to unequal space on multiple disks being used to create a simple volume set, which is not RAID 0. RAID 0 is not RAID in every sense because it doesn’t provide the fault tolerance implied by the redundant component of the name. Data is written across multiple drives, so one drive can be reading or writing while another drive’s read-write head is moving. This makes for faster data access. However, if any one of the drives fails, all content is lost. Some form of redundancy or fault tolerance should be used in concert with RAID 0.

RAID 1 Also known as disk mirroring. RAID 1 is a method of producing fault tolerance by writing all data simultaneously to two separate drives. If one drive fails, the other contains all of the data, and it will become the primary drive. However, disk mirroring doesn’t help access speed, and the cost is double that of a single drive. If a separate host adapter is used for the second drive, the term duplexing is attributed to RAID 1. Only two drives can be used in a RAID 1 array.

RAID 5 Combines the benefits of both RAID 0 and RAID 1, creating a redundant striped volume set. Unlike RAID 1, however, RAID 5 does not employ mirroring for redundancy. Each stripe places data on n-1 disks, and parity computed from the data is placed on the remaining disk. The parity is interleaved across all of the drives in the array so that neighboring stripes have parity on different disks. If one drive fails, the parity information for the stripes that lost data can be used with the remaining data from the working drives to derive what was on the failed drive and to rebuild the set once the drive is replaced.

The same process is used to continue to serve client requests until the drive can be replaced. This process can result in a noticeable performance decrease, one that is predictable because all drives contain the same amount of data and parity. Furthermore, the loss of an additional drive results in a catastrophic loss of all data in the array. Note that while live requests are served before the array is rebuilt, nothing needs to be computed for stripes that lost their parity. Recomputing parity for these stripes is required only when rebuilding the array. A minimum of three drives is required for RAID 5. The equivalent of one drive is lost for redundancy. The more drives in the array, the less of a percentage this single disk represents.

Although there are other implementations of RAID, such as RAID 3 and RAID 4, which place all parity on a single drive, resulting in varying performance changes upon drive loss, the three detailed here are by far the most prolific. RAID 6 is somewhat popular as well because it is essentially RAID 5 with the ability to lose two disks and still function. RAID 6 uses the equivalent of two parity disks as it stripes its data and distributed parity blocks across all disks in a fashion similar to that of RAID 5. A minimum of four disks is required to make a RAID 6 array.

There are also nested or hybrid implementations, such as RAID 10 (also known as RAID 1+0), which adds fault tolerance to RAID 0 through the RAID 1 mirroring of each disk in the RAID 0 striped set. Its inverse, known as RAID 0+1, mirrors a complete striped set to another striped set just like it. Both of these implementations require a minimum of four drives and, because of the RAID 1 component, use half of your purchased storage space for mirroring.

Tape Backup Devices

An older form of removable storage is the tape backup. Tape backup devices can be installed internally or externally and use either a digital or analog magnetic tape medium instead of disks for storage. They hold much more data than any other medium, but they are also much slower. They are primarily used for batch archival storage, not interactive storage.

With hard disks, it’s not a matter of “if they fail”; it’s “when they fail.” So you must back up the information onto some other storage medium. Tape backup devices were once the most common choice in larger enterprises and networks because they were able to hold the most data and were the most reliable over the long term. Today, however, tape backup systems are seeing competition from writable and rewritable optical discs, which continue to advance in technology and size. Nevertheless, when an enterprise needs to back up large amounts of data on a regular basis, some form of tape media is the most popular choice. Table 2.3 lists the best known tape formats in order of market release dates, oldest first, and their capacities. Note that capacities are not associated with the format names but instead with models of tape within each format family.

Flash Memory

Once only for primary memory usage, the same components that sit on your motherboard as RAM can be found in various physical sizes and quantities among today’s solid-state storage solutions. These include older removable and nonremovable flash memory mechanisms, Secure Digital (SD) and other memory cards, and USB flash drives. Each of these technologies has the potential to store reliably a staggering amount of information in a minute form factor. Manufacturers are using innovative packaging for some of these products to provide convenient transport options (such as keychain attachments) to users. Additionally, recall the SSD alternatives to magnetic hard drives mentioned earlier in this chapter.

For many years, modules and PC Card devices known as flash memory have offered low- to mid-capacity storage for devices. The name comes from the concept of easily being able to use electricity to alter the contents of the memory instantly. The original flash memory is still used in devices that require a nonvolatile means of storing critical data and code often used in booting the device, such as routers and switches.

For example, Cisco Systems uses flash memory in various forms to store its Internetwork Operating System (IOS), which is accessed from flash during boot-up and, in certain cases, throughout operation uptime and therefore during an administrator’s configuration sessions. Lesser models store the IOS in compressed form in the flash memory device and then decompress the IOS into RAM, where it is used during configuration and operation. In this case, the flash memory is not accessed again after the boot-up process is complete, unless its contents are being changed, as in an IOS upgrade. Certain devices use externally removable PC Card technology as flash memory for similar purposes.

The following sections explain a bit more about today’s most popular forms of flash memory, memory cards, and USB flash drives.

SD and Other Memory Cards

Today’s smaller devices require some form of removable solid-state memory that can be used for temporary and permanent storage of digital information. Gone are the days of using microfloppies in your digital camera. Even the most popular video-camera media, such as mini-DVD and HDD, are giving way to solid-state multi-gigabyte models. These more modern electronics, as well as most contemporary digital still cameras, already use some form of removable memory card to store still images permanently or until they can be copied off or printed out. Of these, the Secure Digital (SD) format has emerged as the preeminent leader of the pack, which includes the older MultiMediaCard (MMC) format on which SD is based. Both of these cards measure 32mm by 24mm, and slots that receive them are often marked for both. The SD card is slightly thicker than the MMC and has a write-protect notch (and often a switch to open and close the notch), unlike MMC.

Even smaller devices, such as mobile phones, have an SD solution for them. One of these products, known as miniSD, is slightly thinner than SD and measures 21.5mm by 20mm. The other, microSD, is thinner yet and only 15mm by 11mm. Both of these reduced formats have adapters allowing them to be used in standard SD slots. Figure 2.11 is a photo of an SD card and a microSD card next to a ruler based on inches.

Table 2.4 lists additional memory card formats, the slots for some of which can be seen in the images that follow the table.

Format	Dimensions	Details	Year Introduced
CompactFlash (CF)	36mm by 43mm	Type I and Type II variants; Type II used by IBM for Microdrive	1994
xD-Picture Card	20mm by 25mm	Used primarily in digital cameras	2002

Figure 2.12 shows the memory-card slots of an HP PhotoSmart printer, which is capable of reading these devices and printing from them directly or creating a drive letter for access to the contents over its USB connection to the computer. Clockwise from the upper left, these slots accommodate CF/Microdrive, SmartMedia, Memory Stick (bottom right), and MMC/SD. The industry provides almost any adapter or converter to allow the various formats to work together.

Many other devices exist for allowing access to memory cards. For example, 3½″ form factor devices can be purchased and installed in a standard front-access drive bay. One such device is shown in Figure 2.13. External card readers connected via USB, such as the one shown in Figure 2.14 (front first, then back), are widely available in many different configurations.

Many of today’s laptops have built-in memory card slots, such as the one shown in Figure 2.15.

USB Flash Drives

USB flash drives are incredibly versatile and convenient devices that allow you to store large quantities of information in a very small form factor. Many such devices are merely extensions of the host’s USB connector, extending out from the interface but adding little to its width, making them easy to transport, whether in a pocket or laptop bag. Figure 2.16 illustrates an example of one of these components and its relative size.

USB flash drives capitalize on the versatility of the USB interface, taking advantage of the Plug and Play feature and the physical connector strength. Upon insertion, these devices announce themselves to Windows File Explorer as removable drives, and they show up in the Explorer window with a drive letter. This software interface allows for drag-and-drop copying and most of the other Explorer functions performed on standard drives. Note that you might have to use the Disk Management utility (discussed in Chapter 13, “Operating System Basics”) to assign a drive letter manually to a USB flash drive if it fails to acquire one itself. This can happen in certain cases, such as when the previous letter assigned to the drive has been taken by another device in the USB flash drive’s absence.

Hot-Swappable Devices

Many of the removable storage devices mentioned are hot swappable. This means that you can insert and remove the device with the system powered on. Most USB-attached devices without a file system fall into this category. Non-hot-swappable devices, in contrast, either cannot have the system’s power applied when they are inserted or removed or have some sort of additional conditions for their insertion or removal. One subset is occasionally referred to as cold swappable, the other as warm swappable. The system power must be off before you can insert or remove cold-swappable devices. An example of a cold-swappable device is anything connected to the PS/2-style mini-DIN connector, such as a keyboard or mouse. Insertion with the power on generally results in lack of recognition for the device and might damage the motherboard. AT keyboards and the full-sized DIN connector have the same restriction.

Warm-swappable devices include USB flash drives and external drives that have a file system. Windows and other operating systems tend to leave files open while accessing them and write cached changes to them at a later time, based on the algorithm in use by the software. Removing such a device without using the Safely Remove Hardware utility can result in data loss. However, after stopping the device with the utility, you can remove it without powering down the system, hence the warm component of the category’s name. These are officially hot-swappable devices.

RAID systems benefit from devices and bays with a single connector that contains both power and data connections instead of two separate connectors. This is known as Single Connector Attachment (SCA). SCA interfaces have ground leads that are longer than the power leads so that they make contact first and lose contact last. SATA power connectors are designed in a similar fashion for the same purpose. This arrangement ensures that no power leads make contact without their singular ground leads, which would often result in damage to the drive. Drives based on SCA are hot swappable. RAID systems that have to be taken offline before drives are changed out, but the system power can remain on, are examples of warm-swappable systems.

Removing Storage Devices

Removing any component is frequently easier than installing the same part. Consider the fact that most people could destroy a house, perhaps not safely enough to ensure their well-being, but they don’t have to know the intricacies of construction to start smashing away. On the other hand, very few people are capable of building a house. Similarly, many could figure out how to remove a storage device, as long as they can get into the case to begin with, but only a few could start from scratch and successfully install one without tutelage.

In Exercise 2.1, you’ll remove an internal storage device.

Installing Storage Devices

An obvious difference among storage devices is their form factor. This is the term used to describe the physical dimensions of a storage device. Form factors commonly have the following characteristics:

• 3½″ wide vs. 5¼″ wide
• Half height vs. full height vs. 1″ high and more
• Any of the laptop specialty form factors

You will need to determine whether you have an open bay in the chassis to accommodate the form factor of the storage device that you want to install. Adapters exist that allow a device of small size to fit into a larger bay. For obvious reasons, the converse is not also true.

In Exercise 2.2, you’ll install an internal storage device.

Classic Power Connectors

The classic connectors comprise outdated connectors as well as connectors still in use today despite being found in the original IBM PC.

AT System Connector

The original power connectors attached to the early PC motherboards were known collectively as the AT system connector. There are two six-wire connectors, labeled P8 and P9, as shown in Figure 2.18. They connect to an AT-style motherboard and deliver the power that feeds the electronic components on it. These connectors have small tabs on them that interlock with tabs on the motherboard’s receptacle.

The P8 and P9 connectors must be installed correctly or you will damage the motherboard and possibly other components. To do this (on standard systems), place the connectors side by side with their black wires together, and then push the connectors together or separately onto the 12-pin receptacle on the motherboard. Although there is keying on these connectors, they both use the exact same keying structure. In other words, they can still be swapped with one another and inserted. When the black ground leads are placed together when the connectors are side by side, it is not possible to flip the pair 180 degrees and still insert the two connectors without physically defeating the keying. Most technicians would give up and figure out their mistake before any damage occurs if they always place the grounds together in the middle.

It is important to note that only legacy computers with AT and baby AT motherboards use this type of power connector.

Standard Peripheral Power Connector

The standard peripheral power connector is generally used to power different types of internal disk drives. This type of connector is also called a Molex connector. Figure 2.19 shows an example of a standard peripheral power connector. This power connector, though larger than the floppy drive power connector, uses the same wiring color code scheme as the floppy drive connector, although with a heavier gauge of wire. The added copper is for the additional current drawn by most devices that call for the Molex interface.

Modern Power Connectors

Modern components have exceeded the capabilities of some of the original power supply connectors. The Molex peripheral connector remains, but the P8/P9 motherboard connectors have been consolidated and augmented, and additional connectors have sprung up.

ATX, ATX12V, and EPS12V Connectors

With ATX motherboards came a new, single connector from the power supply. PCI Express has power requirements that even this connector could not satisfy, leading to different connectors with different versions of the more advanced ATX12V specifications, which have gone through four 1.x versions and already five 2.x versions. Throughout the versions of ATX12V, additional 4-, 6-, and 8-pin connectors supply power to components of the motherboard and its peripherals—such as network interfaces, PCIe cards, specialty server components, and the CPU itself—that require a +12V supply in addition to the +12V of the standard ATX connector. These additional connectors follow the ATX12V and EPS12V standards. The ATX connector was further expanded by an additional four pins in ATX12V 2.0.

The original ATX system connector (also known as the ATX motherboard power connector) feeds an ATX motherboard. It provides the six voltages required, plus it delivers them all through one connector: a single 20-pin connector. This connector is much easier to work with than the dual connectors of the AT power supply. Figure 2.20 shows an example of an ATX system connector.

When the Pentium 4 processor was introduced, it required much more power than previous CPU models. Power measured in watts is a multiplicative function of voltage and current. To keep the voltage low meant that amperage would have to increase, but it wasn’t feasible to supply such current from the power supply itself. Instead, it was decided to deliver 12V at lower amperage to a voltage regulator module (VRM) near the CPU. The higher current at a lower voltage was possible at that shorter distance from the CPU.

As a result of this shift, motherboard and power supply manufacturers needed to get this more varied power to the system board. The solution was the ATX12V 1.0 standard, which added two supplemental connectors. One was a single 6-pin auxiliary connector similar to the P8/P9 AT connectors that supplied additional +3.3V and +5V leads and their grounds. The other was a 4-pin square mini-version of the ATX connector, referred to as a P4 (for the processor that first required them) connector, which supplied two +12V leads and their grounds. EPS12V uses an 8-pin version, called the processor power connector, which doubles the P4’s function with four +12V leads and four grounds. Figure 2.21 illustrates the P4 connector. The 8-pin processor power connector is similar but has two rows of 4 and, despite its uncanny resemblance, is keyed differently from the 8-pin PCIe power connector to be discussed shortly.

For servers and more advanced ATX motherboards that include PCIe slots, the 20-pin system connector proved inadequate. This led to the ATX12V 2.0 standard and the even higher-end EPS12V standard for servers. These specifications call for a 24-pin connector that adds further positive voltage leads directly to the system connector. The 24-pin connector looks like a larger version of the 20-pin connector. The corresponding pins of the 24-pin motherboard header are actually keyed to accept the 20-pin connector. Adapters are available if you find yourself with the wrong combination of motherboard and power supply. Many power supplies feature a 20-pin connector that snaps together with a separate 4-pin portion for flexibility, called a 20+4 connector, which can be seen in Figure 2.22. The 6-pin auxiliary connector disappeared with the ATX12V 2.0 specification and was never part of the EPS12V standard.

ATX12V 2.1 introduced a different 6-pin connector, which was shaped more like the P4 connector than the P8/P9-style auxiliary connector from the 1.x standards (see Figure 2.23). This 6-pin connector was specifically designed to give additional dedicated power to the PCIe adapters that required it. It provided a 75W power source to such devices.

ATX12V 2.2 replaced the 75W 6-pin connector with a 150W 8-pin connector, as shown in Figure 2.24. The plastic bridge between the top two pins on the left side in the photo keeps installers from inserting the connector into the EPS12V processor power header but clears the notched connector of a PCIe adapter. The individual pin keying should avoid this issue, but a heavy-handed installer could defeat that. The bridge also keeps the connector from inserting into a 6-pin PCIe header, which has identically keyed corresponding pins.

SATA Power Connectors

SATA drives arrived on the market with their own power requirements in addition to their new data interfaces. Refer back to Figure 2.10 and imagine a larger but similar connector for power. You get the 15-pin SATA power connector, a variant of which is shown in Figure 2.25. The fully pinned connector is made up of three +3.3V, three +5V, and three +12V leads interleaved with two sets of three ground leads. Each of the five sets of three common pins is supplied by one of five single conductors coming from the power supply. The same colors are generally used for the conductors as with the Molex and Berg connectors. When the optional 3.3V lead is supplied, it is standard to see it delivered on an orange conductor.

Note that in Figure 2.25, the first three pins are missing. These correspond to the 3.3V pins, which are not supplied by this connector. This configuration works fine and alludes to the SATA drives’ ability to accept Molex connectors or adapters attached to Molex connectors, thus working without the optional 3.3V lead.