RAID Storage--High-level Impact

No longer a boring storage selection, today's RAID solutions are showing RAID, too, can be a glitzy technology with complex, surprising powers.

BY MARIE RANOIA ALONSO

Since an explosion of digital connectivity, asset management and electronic prepress technologies hit the printing industry like a Hollywood blockbuster comet, terms such as megabytes, gigabytes, terabytes and now even petabytes—all used to describe the size of digital storage needs—have filtered into the printing atmosphere.

A logical phenomena.

In the ever-dawning age of electronic prepress, functions such as RIPing a massive file, digitally archiving a file, or executing production of that file through applications like OPI and CTP can take up impressive volumes of digital storage space.

Purchasing one or a teaming of digital storage solutions to master these digital prepress applications is a mission-critical decision all on its own.

One such solution continues to be RAID (Redundant Array of Independent Disks) technology, which is based on multiple disk drives connected through an array controller. RAID systems permit digital data to be distributed across multiple physical drives with redundant disk capacity storing parity information to ensure data integrity and recovery in case of individual disk failure.

"Purchasing RAID storage, in particular, has become a strategic decision," reports Susan Leenerman, marketing director at Ciprico. "It's not just about the amount of storage you need. Each type of RAID has its benefits and its risks. RAID storage must deliver data as fast as possible and continue running at full speed when there is a drive or system failure."

So then, how best to maximize RAID and minimize the potential for lost digital assets? Leenerman offers the following tips.

• Start with a thorough analysis of your own particular applications needs, looking closely at performance and capacity.
• Think about what sort of redundancy features you might want. Redundancy is what RAID is all about. If a prepress operation is cost-sensitive, at a minimum, the RAID system needs to provide a parity drive. The disk drive is one of the most common components in a RAID system to fail. If a prepress facility can spend a little more, it can obtain a RAID box with hot-swap drives and power supplies to ensure that productivity is always up and running.
• Determine the order of importance of the performance, capacity and redundancy. This, of course, depends on the prepress application. Performance measures the speed of transferring and receiving data to the server or desktop system. It can be measured by megabytes per second or by improved productivity. It varies from application and system configurations.
• Choosing RAID technology requires an understanding of the different levels of RAID, known as RAID 0, 1, 3 and 5. Each RAID level offers different degrees of performance, availability and cost.

"While the word 'level' is commonly used with RAID, the term does not mean that the different levels indicate a quality or performance ranking," Leenerman says. "The numbers only indicate different methodologies, all of which are useful for particular types of data or applications."

The Ciprico 7000 series comes with a Fibre Channel interface allowing up to 100MB/second transfer speed. Ciprico's 6900 series has a 40MB/second interface and the same redundancy features. Recently, Ciprico introduced its FibreSTORE disk array for entry-level requirements.

"The trend in the graphic arts to embrace RAID storage technology answers two requirements: Deliver the higher throughput needed to sustain the processing of imagesetters, proofers and applications that could not perform with the limitation of a single hard drive performance, and secure the work during production and the waiting cycles needed between stages," reports Gideon Moran of CyberStorage Systems.

"Most implementations of RAID level 0 will be on workstations doing production or outputting to proofers, imagesetters and direct-to-plate. These systems usually consist of two or four disk drives with either software or a mix of hardware or software products," Moran suggests. "RAID 3 or 4 is implemented with large configurations consisting of multiple drives from five to over 100, with capacities ranging from 10GB to over 1.5TB." Currently, CyberStorage offers

CyberBorg, a RAID Storage Management System that delivers up to 117MB/second of sequential transfer.

At Storage Computer, the focus is the OmniRAID cluster array for Windows NT. The product is designed for commercial printers and prepress shops looking to migrate core business applications from mainframe UNIX servers to NT, without incurring any additional storage management overhead or performance degradation.

Sounds familiar, but the neat thing about this product is that the OmniRAID cluster array automates data protection on a transaction vs. application level, eliminating the need to manage RAID levels continually for different applications stored on multiple NT or UNIX servers. Capacity may be scaled from 54GB to 1TB to enable clustered NT server-based infrastructures to benefit from mainframe-class storage services.

"The demand for simplified, centralized management, the need for enterprise-wide scalability and the realization that software-based storage platform-level functionality enhancements do not consume server or network resources are clearly bolstering the popularity of RAID," adds Anton Murphy, manager of worldwide strategic alliances at Storage Computer.

MicroNet Technology—acquired by Ampex Corp. earlier this year—puts its RAID energies into DataDock 7000, a high-capacity, fault tolerant, data storage solution featuring RAID 0, 1, 3, 5 and 0+1.

DataDock 7000 offers auto-configuration and MicroNet's configuration assistant, which reduces technical support requirements. DataDock 7000 is compatible with NetWare, Windows NT, UNIX and Mac OS, with hardware supported by IBM, Compaq, Silicon Graphics, DEC and virtually any CPU using SCSI interface.

Not unlike many RAID towers, DataDock can seem intimidating, with seven drive bays per DataDock 7000, hundreds of gigabytes on a single SCSI bus, Fast SCSI-2, Fast/Wide SCSI-2, Ultra and Ultra/Wide DataDock 350 modules.

Sometimes, intimidation is good.

DataDock 7000's design eliminates interdependence between components, providing nonstop operation, even in the event (gulp!) of component failure.

More commercial printing and prepress firms are moving to central storage, states Brian Lee, prepress business development manager at MicroNet. "Capacities are becoming higher and higher; RAID levels are proving to be a strong support for high capacity needs."

Lee contends today's current and emerging local area networks (LANs), while adequate for many tasks, were simply not designed to handle large digital files, sometimes more than 100MB or larger. Storage Area Network (SAN) technology, based on Fibre Channel technology, is proving, and will continue to prove, itself the hot ticket interconnection model for providing scalable, high-speed network connections for data storage.

As for Ciprico? Leenerman offers a parting tip for those considering a move to the multiple levels of RAID technology. "If you're going to invest money in a robust RAID system, you should also invest in a high-performing server," she advises. "The server, after all, is the cornerstone."

RAID 101:
Lessons in Storage Levels

RAID level 0
Uses a method of writing data to a group of drives called striping. The benefit here is more performance than one would get from a single drive. The risk is that RAID 0 provides no redundancy. No redundancy might be a problem when a prepress operation needs to meet ongoing, rigorous demands.

RAID level 1
Drives are mirrored or duplicated. This RAID architecture automatically copies data on a second set of drives, creating an exact copy of all files. This offers complete redundancy but is an expensive solution since you need to buy twice as many drives.

RAID level 3
A high-bandwidth solution for applications that transfer large sequential files such as OPI, RIP and CTP. The data are written in bytes across multiple drives with one drive holding the parity or error-correcting code. Writing at the byte level enables calculations to be made on the fly, without having to first read data into a cache. There is no performance degradation. RAID level 3 offers full redundancy.

RAID level 5
A transactional solution. This level of RAID works well with database applications because its implementation offers higher performance in small request sizes. Data are written in blocks across all drives with parity distributed among all the drives. But when a drive failure occurs, there is performance degradation. RAID level 5 also offers full redundancy.