Archiving made easier

Fast disk-based media make great alternatives to tape.

When Pope John Paul II died in April, editors at The Dallas Morning News quickly pulled together a photo retrospective of the pontiff's life - a project they would have considered impractical not long ago.

The problem of old was the time and effort it took to get images from the tape archives - a task only IT could do. But last summer, Bob Mason, director of publishing systems, moved the paper's filmstrips and digital pictures to a new disk-based archival system that gives editors speedy access to stored data.

The Dallas Morning News is struggling to get a handle on the value of its business data, then to craft a tiered-storage strategy around that understanding. As it has for Mason, archiving is bubbling up as an increasingly important part of that storage decision.

"We don't want to build a two-to-three-year solution but one that will last for 20 years," Mason says.

Adopting a tiered storage and archival process means determining how retrievable and available data needs to be, says Jim Geis, director of storage solutions for Forsythe Solutions Group, an IT consultancy. "Just because some information is old, static or considered reference information, it still could be mission-critical," he says. "You may not need to access the data very quickly, you may not have to have it constantly available, but you still need to have it available .... [with] high integrity."

Business-critical applications such as ERP or Oracle database applications typically need the most highly available, fastest Fibre Channel storage, while imaging applications tend to require less quickly accessible, less expensive storage such as Serial Advanced Technology Attachment (SATA) or Ultra Density Optical (UDO ). Tape is hardly an option for data that needs to be accessed even occasionally.

For The Dallas Morning News, Mason investigated a variety of archival media - magneto-optical, SATA and even EMC's near-line, ATA-based Centera system for content-addressable storage - before settling on a disk-based system. That system, Plasmon's G638 optical library with a 19T-byte capacity, supports 30G-byte, write-once UDO media cartridges.

UDO products, which have been available for about two years, have a 5.25-inch form factor, making them suitable for use in tape libraries. Besides Plasmon, which helped define the specification, IBM and HP have adopted UDO technology.

Over the next three years, UDO capacity is expected to double to 60G bytes, and then again to 120G bytes.

"We have to have a media that can grow with the increasing resolution of digital cameras," Mason says. Scalability is a huge issue when you're storing some 2 million photographs a year, he adds.

The lasting nature of data stored on UDO media factored in to Mason's choice, too. "For our digital archives, we have to have permanence and authenticity - we want to make sure we write data once and never have a risk of losing that data," he says.

Wolfgang Schlichting, an IDC research manager, agrees that UDO is a great tape replacement. "Due to its compelling cost per gigabyte and write-once support, [we] believe UDO will quickly replace magneto-optical in storage solutions," he says.

A good IDE-a

But UDO is far from the only choice for archiving data in the new data center.

At Good Samaritan Community Healthcare, in Puyallup, Wash., simple and inexpensive Integrated Device Electronics (IDE) drives - the kind of drives used most often in desktop PCs - fit the bill. The organization archives more than 35,000 patient radiology studies a year on IDE drives in Permabit's Permeon Compliance Store system, says Eric Lowe, IT and operations manager at Good Samaritan.

Lowe simultaneously stores new radiology studies on the hospital's Picture Archive Communication System (PACS) and the IDE drives. The PACS sits on a storage-area network (SAN ), which comprises fast and efficient, but expensive, Fibre Channel drives. After six to nine months, Lowe purges the studies off the Fibre Channel SAN but they remain on the Permabit system for a time ranging from 10 years to the lifetime of the patient, Lowe says.

"Doctors don't know whether the study is coming off the high-speed Fibre Channel SAN or the not-quite-so-high-speed Permabit appliance," Lowe says. "Our goal was to make it transparent."

In choosing Permabit's IDE-based appliances for his archiving needs, Lowe says he compared the cost of the media with the value of the data. "I couldn't afford to build another Fibre Channel infrastructure for 10 terabytes of data. I had to find a solution that was going to be cost-effective so I didn't have to keep these images online all the time," he says.

Archiving Choices

Technology

Availability

Performance

Cost

(per gigabyte*)

Vendors

Solid-state disk

High

High

$2,000 to $5,000

SolidData, Texas Memory Systems

Fibre Channel

High

High

$40 to $70

EMC, HP, IBM

Serial ATA

Moderate

Moderate

$15 to $35

EMC, LeftHand Networks, Nexsan Technologies, Compellent Technologies

Ultra Density Optical

Moderate

Moderate

$2

HP, IBM, Plasmon

Tape

Low

Low

50 cents to $3

ADIC, Spectralogic, StorageTek

*SOURCE: HORISON INFORMATION STRATEGIES

And like Mason, Lowe rejected magnetic tape for archiving because of its inability to access data fast. The hospital uses some magnetic tape for archiving, such as for cardiology, but plans to migrate those images to the Permeon appliance within the next few months, he says. Lowe calls archiving cardiology images on tape a "really bad" idea. You can't make a cardiologist wait two hours for an archived image when he's got "a patient on the table," he says.

Serious about SATA

Hal Weiss, IS engineer for Baptist Memorial Health Care in Memphis, Tenn., chose a compromise between magnetic tape and disk storage for the medical images he needed to archive. He uses EMC's Centera for archiving PACS information but stores other data on a disk-based system from start-up Copan Systems. In its Revolution 200T, Copan uses a unique technology that enables the system to spin disks only when data needs to be pulled off. This reduces power consumption and helps prevent wearing down of the disk platters, Copan says.

Plus, Weiss says, the Copan system can store a large amount of data inexpensively. The hospital stores 15T bytes of PACS images and 6T bytes of patient folders a year, Weiss says.

The Centera and Revolution 200T systems use SATA drives, the faster, easier-to-configure replacement for parallel ATA (or IDE) drives used in many of today's servers and desktop PCs. But adding capacity to the Centera means increasing processor power, where it doesn't for the Copan system, Weiss says. "For 32 terabytes of Centera, I have to use two cabinets. I can get 224 terabytes of Copan storage in one cabinet, and save on expensive floor space."

Even with faster, more efficient options, tape won't disappear entirely as an archiving medium.

At the Public Broadcasting System in Alexandria, Va., for example, tape makes the perfect Tier-4 choice, says Ken Walters, the firm's senior director of enterprise systems. Walters stores the data he never needs to get to on tape. He reserves Tier 1 storage for business-critical data and Tier 2 and 3 storage for the development environment.

"Tiering is always a work in progress," Walters says. "The challenge is getting everything where it belongs, given staff and time issues."

Clearly, IT is tiering storage in response to business pressures and learning how to archive data based on its importance to the business.

Learn more about this topic

SATA reduces storage costs

08/25/03

Serial storage garners big name support

05/17/04

What lasts longer - SATA, serial-attached SCSI or Fibre Channel?

03/21/05

UDO boosts storage capacity

09/27/04

UDO replacing MO faster than expected, says Plasmon

10/13/04

Zetera debuts low-cost storage option

01/24/05

Network World's Storage in the Enterprise newsletter

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:

Copyright © 2005 IDG Communications, Inc.