GET READY FOR INTELLIGENT TAPE.

The Third Wave is almost upon us

The fundamental concept of change remains a constant in the world. The Information Technology (IT) industry is entering the next millennium with the promise of even more profound change and consequences than at any previous time. The Internet, which actually began in 1969 at UCLA's Network Measurement Center with four nodes, grew to 100 nodes by 1975, and had only 5 million users by July 1995. By the end of the year 2002, the number of Internet users worldwide could approach 300 million! Storage density has increased between 60% and 100% every 12 months since 1990, while the price per unit of storage has dropped at about 30% per year throughout the same time period. Storage capacity is no longer a limiting factor to any digital application as storage technology can physically contain any and all applications being developed.

As an industry, we are finally beginning to realize that data is now the DNA or lifeblood of an organization, no matter what its size. Anyone can have the identical server, the same hardware, and most of the same software products. However, no other entity can have your data about your business, your customers, and ultimately produce your information. It is now the data, not the hardware, software, or network that makes an organization unique. It is the transformation of that data to more meaningful constructs, namely information and later knowledge, that will make an organization flourish.

From a storage perspective, the 1990-decade witnessed more technology advancement than there has been in the entire history of the data storage industry. Beginning with a punched card and soon followed by the first magnetic tape drive, the IBM 726 in 1952, and the first disk drive, the IBM Ramac 350 in 1956, the storage industry and its hierarchy was born. Throughout the first 40 years, the primary focus of the IT industry had been the computer platform, not on data or storage. The continued decline in the cost of computing power and storage capacity has enabled many new applications to be developed that were previously cost-prohibitive and left in analog or non machine-readable storage formats.

Software has become the gating factor in time-to-market for most products and the rate of improvement in software development has been extremely slow compared to the rapid pace of technology advances. Software costs are rapidly rising and now represent an estimated 285% of the hardware expenses in a typical IT organization. Storage costs can often account for over half of the total IT hardware budget. With areal density progression and transistor speed well-defined for some years to come, improving software development capability represents one of the biggest upcoming challenges and opportunities for the storage industry in the next millennium.

Today, the new functions that will significantly improve storage capability arrive in the marketplace at the speed of software completion, not at the rate of hardware development. The new storage subsystems will nonetheless have to deal with much higher software content than ever before, thanks to the emerging SAN. Many SAN components, including disk controllers, tape subsystems, and the elements of the SAN fabric, will have processing capability and run real time operating system kernels as storage intelligence becomes more widely dispersed throughout the SAN. Our expectations for software improvement are mounting.

What are some of the great expectations for future storage subsystems? First of all, disk storage is poised for its third generation architecture. The first generation could be described as the JBOD (Just a Bunch Of Disks) directly connected to a server. The second generation was defined as the RAID era and began in the early 1990s. Within the RAID era, RAID first meant Redundant Arrays of Inexpensive Disks. Even with a 30% or greater price erosion annually, RAID became known as Redundant Arrays of Independent Disks, as inexpensive always seemed too expensive. The third generation of disk storage will soon embrace RAID as Redundant Arrays of Intelligent Disks. Inexpensive and low-cost microprocessors make it feasible for the first time to add computing power to the storage sub-system or even the device. This added intelligence would be used to more easily extend the limits of scalability while reducing some of the many bottlenecks in the I/O path. The latter change is long overdue.

In parallel, storage related computer cycles are offloaded from the server, thus making more computing power available to the core applications of the business. Disk drives and subsystems may see the advent of dual-actuators as capacities increase faster than offsetting performance gains. This characteristic is best described as Access Density, the ratio of disk drive performance to its capacity. Access Density has steadily fallen for nearly forty years and ranges close to one I/O per second per GB for most drives currently available. Dual actuators have been examined before but have always proven to be too costly. The re-emergence of the Solid State Disk is on the expectation list. Taking advantage of fault-tolerant designs that are essentially non-volatile, capacities up to 16GB or greater, and prices well below $20 per MB, SSD's bring dramatic performance and response time improvements to the numerous block-level I/O performance issues remaining in database structures.

The evolution and progress of tape subsystems has accelerated in the past year and a longterm vision is emerging. The first generation of tape lasted for 35 years and was defined as the era of manual tape. By the middle 1980s, people were generally dissatisfied with the labor-intensive issues associated with tape storage. The second generation of the tape industry was effectively launched in 1987 with the successful introduction of automated robotic tape libraries from StorageTek. Labeled Nearline, this level of storage defined a new mass storage layer between disk and shelf storage, resolving many of the problems inherent with people physically placing tapes on a tape drive. Ten years later, over 30 companies were delivering tape automation solutions.

The third generation of tape promises to be the era of intelligent tape and the expectations are great. Intelligent tape storage will include SAN ready subsystems with additional management intelligence in the controller. Intelligent disk subsystems and intelligent tape subsystems will, together with software, enable outboard data movement to occur between disk and tape with minimal server intervention. Backup and recovery applications obviously are at the top of the list. Intelligent or smart tape cartridges with built-in Flash memory or EEPROM chips are currently available and should soon take on a much higher profile as a method of intelligent navigation for file and record finding. Software applications' support for the smart cartridge has lagged to date.

Virtual tape capability has arrived though its true capabilities have barely been addressed. Virtual tape is poised to play future roles that were not part of the original virtual tape vision. As an example, virtual tape libraries on a SAN can appear to a server as any of the industry's numerous tape formats without changing existing procedures for a specific tape technology. Borrowing the RAID concept from disk arrays, tape arrays will also play a role for data-transfer intensive applications. Tape's third generation also promises unprecedented tape capacity increases with a much lower price per megabyte purchased for tape storage. The LTO (Linear Tape Open) roadmap calls for a fourth generation version of the Ultrium cartridge containing 800GB native capacity. A typical data compression ratio of three-to-one would yield a 2.4TB capacity, twice the capacity of two of STK's original Silos on a single piece of media. This represents a volumetric efficiency improvement of 12,000 to one.

The accelerated pace of technological progress is causing many of the traditional barriers in the storage industry to fall. Conventional subsystem boundaries are changing, as storage becomes more and more independent of any specific server connection. As processing intelligence expands to all parts of the computer system, the great expectations for the storage industry are now within our grasp.