Computer Storage Architecture - From Ancient Stone Age to Modern Computer Storage

Our ancestors used to carve on stones to record their important events, histories and stories. Then, the Chinese invented paper and methods for mass producing it some three thousand years ago, after getting tired of carving on stones and bamboo. This made storing information much easier and less labor-intensive. Paper has remained throughout the last three thousand years the primary medium for societies to record information.

However, there are many issues related to the use of paper for storing information. For one thing, paper is not very durable. It can also be quite heavy, although not as heavy as stone! But by far the biggest issue with paper is accessibility: finding the information that we need within a document quickly and easily.

Fast forward to today. Since the invention of the computer disk, we have devised three main ways of storing massive amounts of data and information onto disks. These three ways are Direct Attached Storage (DAS), Network Attached Storage (NAS), and Storage Area Network (SAN).

Direct Attached Storage

By my estimate, 95% of all computer storage devices today are DAS. For example, just about all desktops and laptops are using DAS. By saying DAS, we are describing the disk storages that are directly connected to the computers that are using them, either internal to the computers or external to them, like external USB disk drive. Most of the servers today are also using DAS. One of the main reasons why so many computers are using and sharing DAS is because of the design decision made when the computer was first introduced. When computers were first introduced, configuring them together in a networked environment was, in general, not part of the consideration. Computer networking has become pervasive only during the last 15 or so years. Despite the fact that computers were operated for a long time without any DAS (so called diskless workstation), the diskless model never caught on as a popular way of deploying computers.

In some cases, DAS is unavoidable, for example, with laptops: we need the storage to go with us wherever we take out laptop, so not having DAS for the laptops is just not an option. DAS is also ideal for low-cost deployment with relative localized applications.

There are a few disadvantages of using DAS. DAS is a captive storage, generally speaking. Once the storage device is attached to a computer, it is very difficult to redeploy it even when the situation arises where one computer has plenty of storage space and another is running out of space. There is not an easy way to reallocate the storage space to resolve this situation. Since there is not an easy way to predict most of the time how the disk storage is going to be used on a given computer, installing the right amount of DAS is always tricky. With DAS, there are also backup issues; since computers with DAS are distributed, it is next to impossible to back them up easily and quickly. So most companies just back up the servers and never or rarely ever back up the desktops and laptops. The distributed nature of DAS can also make managing and maintaining it a nightmare for Information Technology (IT) professionals.

Network Attached Storage

With the development of Network File System (NFS) by Sun Microsystems, Inc. for UNIX and Common Internet File System (CIFS) for Windows by Microsoft, a new way of storing files became possible. With these widely accepted standards, storing files in storage devices that are directly connected to the computer network and which are specially designed and optimized for file accessing has become quite popular.

NAS solves the issues of needing to predict the disk storage usage for the most part since most NAS implementation allows the expansion of the storage easily and dynamically.  NAS is also relatively easy to deploy and maintain. Since NAS is built independently of the Operating Systems, it can serve any computers that support NFS & CIFS, and just about all the computers support either or both protocols. NAS can also be centrally deployed and managed, and it supports two of the most popular file systems: NT File System (NTFS) and UNIX File System (UFS). As NAS keeps more and more mission-critical data, many fault tolerance, and data protection and expansion features have been added to NAS over the last several years.

Despite its popularity due to its ability to provide a network-based storage, there are several drawbacks to NAS. Computers that are using NAS store their files remotely over the network, so the applications are unable to access data at the block-level. However, accessing data at the block-level is essential for all database and server-based applications. Backup support in general is also slow and inefficient since all the backups have to be performed at the file system level. In some environments, NAS may also have scalability issues.

Storage Area Network

Two main characteristics of SAN are how it operates when it comes to data transfers within its system and between external devices, and how SAN lays out all the hardware to serve the data with super performance and increase redundancy. SAN transfers data in blocks to increase the throughput performance as well as to allow the database and some server-based applications to work seamlessly in their native environment. Unlike NAS, which usually comes as one or two separate units, SAN comes with several different units and other support devices like specialized SAN switches, separate management units, etc.

Just about all SAN units can also serve up NFS & CIFS, just like NAS. However, not all SAN units use NTFS and UFS; most have their own proprietary file system that understands NTFS and UFS. While NAS uses a general purpose network for data transfers, SAN uses its own very highly sophisticated network to perform some of its data transfers via its web of fiber channels, hundreds to thousands of gigabits at a time. Backup in a SAN environment can also be done in blocks to reduce the time and resources necessary to backup mountains of data.

So, it is due to issues of cost, complexity, and lack of common standards that are the three top reasons why SAN has not been as widely deployed as NAS up to this point.

Like all areas of IT, the storage industry is ever-changing. We are seeing convergences in many of the methods described above; and with the introduction and standardization of iSCSI protocol and IP SAN technologies, we are seeing more creative and less costly ways of using these new combinations as well.

As an old IT saying goes, garbage in, garbage out. What is common with the state of all storage technologies is that none of these technologies addresses the issue of managing the content being stored inside.  By managing, I mean that users should be able to find the files they need when they need them; there should not be any duplications anywhere; and yet, different version of the same files can be kept by certain users, who can also perform automatic data migration based on watermarks or retention policies, in order to meet certain governmental requirements.

I am an optimist and I have a hope. The day will come when we all can carry the entire Library of Congress, the library collections of all the major universities, and all the content currently found on the Internet in our handheld devices, and we can search, read, or listen to this content anytime, anywhere. And more importantly, owning such a device will cost us only a price comparable to dinner in a nice restaurant, getting it refreshed monthly via our wireless network will cost no more than what a can of soda costs, and it can be refreshed within the time it takes to drink that very same can of soda! One day…

By Benson Yeung, Senior Partner

Back to Top