The basics behind storage architectures

In 1993, when Intel ushered in the modern computing age by releasing its Pentium chips, devices roared along at then-astounding speeds of around 60 MHz. Today's mainstream CPUs reach speeds of around 3.4GHz, about 56 times faster than the 1993 Pentium processor.

That growth is impressive, until you consider that 2003-vintage PCs often shipped with as little as 40 megabytes of storage. Today, mainstream PCs ship with 500 gigabytes, 1,250 times more storage than a 1993 PC.

It is even possible to acquire a single hard disk containing one terabyte (TB) of storage--2,500 times the capacity of a 1993 drive.

These figures show that storage growth has outpaced even the growth in raw microprocessor speed in the last 15 years.

The reason for the explosion in storage is that there are now many more ways to create data, and many more types of data that warrant storage.

In 1993, e-mail was in its infancy, the notion of storing music as data was fanciful and digital cameras were yet to become affordable.

Many new technologies have emerged since 1993, and their combined effect has greatly inflated the amount of data created every day, and hence, the need for storage.

Several methods for storing data have evolved to cope with the requirement to preserve information. Here are four key storage architectures:

• Server-attached storage
• Network attached storage
• Storage area networks
• Tape libraries

Server-attached storage
Server-attached storage is the most common type of storage and has perhaps become a little unfashionable.

As the name implies, all of the disks and devices used to store data are installed in the server itself, rather than in a separate device. That makes server-attached storage cost effective, especially for small and midsize businesses (SMBs), because they will not need to buy any other device or system.

This technique is more suitable for small servers that support access to files for several PCs or users in a small office network. It is also a reliable way to provide storage for an application that runs on a server.

However, one challenge companies face with server-attached storage is scalability, as a server can only support so many hard drives.

What this means is, if you run out of space and need new storage, the only alternative is to use bigger disks. But, when your company starts using biggest disks, the migration or upgrade path to a bigger storage network becomes more complex.

Network attached storage (NAS)
The primary difference between NAS and server-attached storage is its use of a dedicated appliance to house disk drives. NAS boxes, as they are known, connect to a network and can be accessed by many client devices, such as PCs or servers.

These boxes are now available in various sizes ranging from single drive units costing a couple of hundred dollars each, to million-dollar systems each packed with dozens of disks.

NAS came into prominence in the late 1990s as a way to help servers run applications faster, relieving the server from having to deliver data files to users.

This form of storage architecture also allows companies to keep files and data they need frequently in a dedicated machine. NAS is also popular for its ability to hold multiple disks, which allows the mirroring of data to ensure reliability and data protection.

NAS machines have proliferated into many specialist forms. In micro or small businesses, NAS units are often used as a substitute for a server.

In large businesses, specialist NAS devices speed access to data such as databases or e-mail archives. Larger businesses use NAS as a place to store data they use frequently, rather than generating network traffic that flows through other storage appliances, such as storage area networks.

Storage area network (SAN)
Today's most advanced storage option is a storage area network (SAN). These storage networks can encompass one or more storage devices, but are perceived by applications and servers as a single source of data.

This is important for users who need a lot of storage capacity because by combining devices into a single logical entity, SAN supports the use of different devices with different capabilities to store data of different importance.

These storage networks are also powerful because they allow businesses to better control their storage costs by blending different types of storage hardware in a single logical unit.

Files that are used every day, for example, can be stored on fast, expensive disks so that users can access them in a flash.

Older data that are only accessed once a month can be shuttled off to older, slower and cheaper storage devices. Access to data here is slower but since the files are used less frequently, user expectations can be better managed.

Because all of these storage options reside in a single SAN, which the operating system manages as one single unit, businesses can mix and match storage to achieve higher performance and price ratio.

One important element businesses must keep in mind is that, with SAN, they will need to deploy and manage a network to connect their storage devices.

This network can be implemented based on Fibre Channel, a protocol that is mature and widely deployed and provides faster, more sophisticated access to data. However, this form of network is more expensive to acquire and operate than an Ethernet network.

Fibre Channel is more costly because such network systems sell at much lower volumes than Ethernet. As such, there are fewer trained technicians capable of operating a fiber channel network and this drives up labor and overall costs for such networks.

A rival Ethernet-based standard Internet SCSI (iSCSI), has since emerged, providing speeds approaching 10Gbps, compared to Fibre Channel's 8Gbps.

Whichever platform an organization chooses, it should remember that speed is important if its employees often access larger files--a task for which SANs are ideal because of their access speed capabilities.

Tape and tape libraries
Any of the storage technologies mentioned above are more than capable of creating backups for a company's data files. All can be deployed with Redundant Arrays of Inexpensive Disks (RAID), a technology that allows the integrated use of two or more hard drives so businesses can attain better redundancy, and hence, greater performance and data reliability.

However, long-term data storage on disk is not a common practice because disk storage is still comparatively more expensive than tape. In addition, disks operate inside servers so NASs or SANs consume power whenever they are in use, and this can be a costly environment for SMBs to maintain--as well as an environment that is not eco-friendly for its high energy consumption.

Another argument against the use of disk as backup is its bulkiness.

Tape is, therefore, often advocated as the best way to store backups and long-term archives. Tapes are generally cheaper than disk because they can be stored on the shelf, instead of inside a disk array, and consume less electricity.

Tape also has the advantage of being inert when not in use, which means they break less often than disks.

A dazzling array of tape drives is available in the market today, ranging to top-end tape library systems that automate the storage and retrieval of tapes, making it easy to access the tape--and data--companies need to retrieve.

Accessing data from tape, however, takes a little longer as businesses still need to identify and locate the appropriate part of a tape to retrieve a particular data they want. Disks, by contrast, allow for the retrieval of any file or data within moments.

Simon Sharwood is a freelance IT writer based in Australia.