After punch cards, tape storage might be the most inconvenient way to keep data around. Tapes, though cheap, are awkward to store and hard to handle. The tape heads require regular cleaning, just like those in a VCR. However, tape is still the cheapest storage option for backing up data. But new technology, including networked storage, is finally making disk storage competitive, especially when the management costs are factored in. And when it comes to compliance, disks win hands down.

According to Peter Gerr, an analyst at Enterprise Storage Group, the worldwide capacity of compliant records stored on disks will increase at an a compound annual growth rate of 172 percent. The percentage stored on tape, however, will decrease by 5 percent each year.

"Compliance is quickly emerging as the primary issue for technology vendors and IT and business professionals," he said.

At MidAmerica Bank, tape storage was a big headache for branch managers. The bank, which also offers brokerage services to its customers, used to require staff at its 43 branches to do regular tape backups.

"The branch staff, which was nontechnical, had to make sure the tape was changed weekly and to communicate if there were any problems," said Ray Zamora, the bank's VP of network operations. "This became very difficult."

Trying to talk a user through basic diagnostics and maintenance over the telephone was a chore, Zamora remembers.

"We were expecting nontechnical people to help with technical problems," he said. "Our backups would fail because our drives were dirty."

Each branch had to spend $7000 to $10,000 for tape drives, software and initial setup, as well as the repeated cost of media, maintenance, communications and transport. And for all this money, the backups weren't perfect-since the backups were performed once a day, the data on the tapes could be several hours old.

This fall, the bank decided to replace its network of tape drives with a Dell EMC storage area network (SAN) running double-take backup software from NSI Software. NSI, which counts 10 of the top 11 securities companies on the Fortune 500 list among its customers, is based in Hoboken, N.J.

Within a week, Zamora expects to have automatic backups running for both the company's mainframe and all of its remote locations.

All of the live data from branch servers are backed up in the background to a central data location. To save on bandwidth, applications and other nonchanging files are not duplicated, and the backup process takes place around other traffic flow. That means the backed-up data could be a few seconds or a couple of minutes old-but this is a big advance over the several hours of lag on a tape backup.

Zamora is also building a second data center, connected to a first by a gigabit circuit. Meanwhile, he still uses tapes to back up the main data center, but that's a much more streamlined process than doing it at 43 remote locations. In fact, Zamora estimates that the bank will save about $50,000 annually by eliminating the need for technical maintenance and assistance.

He added that he will continue to expand on his Dell EMC SAN systems, adding about a terabyte of storage a year. For example, in addition to the growth of the bank itself, and the attendant natural increase in data, he expects to install a new imaging system that will require a lot of storage, he said.

The only problem with the SAN system that he's noticed is that it works too efficiently, he said. If an employee accidentally deletes a file, the change is mirrored to the backup site in seconds.

"Then we go back to tape and it takes 30 minutes to restore," he said. "That's the only negative I can think of-it updates too quickly."

Speed was also an issue for Costa Mesa, Calif.-based CNA Trust, one of the nation's leading retirement plan providers. By mid-2002, CNA Trust had over 300 gigabytes of data that needed to be backed up. The recovery time from a tape backup was over 36 hours, according to John Evans, CNA's VP of network operations.

Today, the company is replicating 60 gigabytes every 36 hours to backup storage servers in Los Angeles, using NSI's Double-Take software. A full recovery now takes only four hours, including all critical systems and Web presence, said Evans.

Worms on Disks

One problem with disk storage, at least from a Wall Street compliance perspective, is that it's not permanent enough-it's too easy to change data and too easy to delete data.

EMC has recently come up with a way around this problem. The company's Centera storage system, released last year, replaces the traditional file-name approach to saving information with digital fingerprints.

When a record is stored, a digital identifier is generated for the record. That identifier is then used instead of a file name to keep track of where the record is. A different record can't be written over the first one because they would have different identifiers-even if they only vary by a bit.

"You can actually demonstrate when you recover objects that what you've got is a true and effective original," said Roy Sanford, VP of content addressed storage at EMC Corp.

The Centera technology can also be used to set expiration dates for records. Attempts to delete information before that date are not allowed, and the data can be deleted automatically at the expiration date using military-grade virtual shredding technology.

"It's the first disk-based Worm technology on the market today," Sanford said.

Worm stands for "write once, read many" and usually refers to optical disk storage, which can be written only once. A more common consumer example is a read-only CD-it can be played many times but only burned once.

"It's a perfect fit for some highly regulated industries like the broker-dealer community," said Sanford.

As a result, he said, six of the top 10 broker-dealers have already installed Centera in their compliance storage infrastructure.

For example, Raymond James Financial, based in St. Petersburg, Fla., bought six terabytes of Centera storage-enough to store the equivalent of 500 billion typed sheets of paper-largely to archive the company's e-mail and document images.

"If lawyers wanted, say, all e-mails sent by a certain person containing a certain keyword, depending on the length of time involved and the number of messages sent, it could take several weeks on older storage systems that use jukeboxes of optical disks," said Karl Schoellnast, VP of IT operations and networking at Raymond James. "Because data will be stored on magnetic disk in the new EMC system, retrieval times should be greatly reduced."

EMC's Centera is a little more expensive to buy than optical storage, but is cheaper in a three-year period when total cost of ownership is taken into account, Sanford said.

It typically takes between one and three days to make the changes for a simple integration between the Centera storage system and a database application, Sanford said, though many major vendors already support it.

Scottrade, a leading online brokerage firm, recently purchased 10 terabytes of EMC Centera Compliance Edition to handle primarily e-mail storage but other types of records as well.

"Because Centera is integrated with most major records management applications, we have the option to use the solution as an enterprisewide online repository for other electronic records, which could include electronic documents, check images and transaction records," said Jeff Polsgrove, CIO at Scottrade, in a statement.

In the first 14 months that this technology has been available, Sanford said, EMC has sold more than 6.5 petabytes of Centera storage to more than 300 customers. By comparison, the entire U.S. Library of Congress takes up about two petabytes. And early indications are that Centera is as good as optical disk storage for compliance with recent Securities and Exchange Commission regulations concerning data and communications storage.

"The SEC does not give a good housekeeping seal of approval," said Sanford. "But a number of our customers have written their 90-day letters to the SEC and they've already gone through that 90-day waiting period, which is why six of the top 10 are already running Centera in their compliance infrastructure."

Microsoft Moves In

There's also been another major change in the networked storage arena. Until a couple of years ago, networked storage servers had been running proprietary operating systems from vendors such as EMC. Then Microsoft entered the market. The latest version of its storage operating system, Microsoft Windows Storage Server 2003, came out in September.

In 2001, Microsoft immediately grabbed 20 percent of the storage operating system market by unit shipments, said Gartner analyst Pushan Rinnen. By the middle of 2003, that number was up to 40 percent. And that growth has been slowed by the fact that vendors were waiting for the fall release of the latest version of the operating system, she added.

"A boutique operating system that doesn't really relate to the rest of the universe isn't enough to win anymore," said Tom Joyce, senior director of NAS product marketing at EMC. "Customers want something that behaves well and works like Windows-or is Windows."

As a result, even EMC has begun offering the Windows operating system on its low-end storage servers.

"We want to be able to sell our array into the customer environment, whether they want Windows or our own operating system," Joyce said.