Banking on Data Replication
Banking on Data Replication
In South Florida, hurricanes are an expected and predictable emergency. Homeowners shutter windows, companies back up data and send it off-site before a storm strikes, and afterward life and business resume as usual.
Such was the contingency plan for Banco Santander Central Hispano International, a Spanish financial institution with U.S. headquarters in Miami, which for six years made do with daily tape backups. Unfortunately, since Sept. 11, 2001, hurricanes arent the only fear in South Florida.
"With the advent of terrorism, we decided we might not get the chance of having that many days of warning that something may happen," said Augustin Abalo, vice president and CIO of Banco Santander and president of the Florida International Bankers Association.
Tape backup may be fine for most businesses, but for a worldwide bank that handles $1 billion in transactions a day, an hour of downtime could mean a loss in the tens of thousands of dollars.
In addition, failure to meet data restoration requirements brings about intangible costs such as the risk to the banks reputation and the risk of violating federal guidelines. "We needed to be able to immediately continue processing even if we had to abandon our premises," Abalo said.
Abalos concern coincided with new guidelines created by the Securities and Exchange Commission. To strengthen the resiliency of the financial services industry, the SEC strongly recommends that banks distance their primary places of business from contingency sites and ensure that personnel are available should the primary sites go down.
Two years ago, Banco Santander went looking for an application that could make continuous processing available at both its production site (Miami) and contingency site (New York). Banco Santander wanted to eliminate the need to courier tapes as well as the daily intervention of an engineer.
Prior to 1999, SANs (storage area networks) were the only game in town for true data replication. But SANs often came with limitations, such as vendor lock-in for both primary and secondary sites.
For Banco Santander, constrained by its budget and connected by a WAN to its contingency site 1,300 miles away, a SAN was not an option. It was costly to deploy and required a dedicated connection and a physically close secondary site.
In addition, synchronous replication over a WAN would hurt the performance of financial processing applications.
Along came Veritas Software Corp.s server-based replication solution, Volume Replicator, which eliminated many constraints inherent in SANs. With Volume Replicator, data could be copied across dissimilar devices from different vendors. A dedicated connection was not required, since data could traverse over IP. Today, forced proprietary protocols are no longer a requirement for SANs.
Most beneficial for Banco Santander was Volume Replicators ability to quickly duplicate data at great distances over the companys WAN. This was possible due to the use of asynchronous replication. Unlike synchronous replication, asynchronous replication copies data to the second source without freezing the application at the primary location.
Volume Replicator became a wake-up call for the SAN market.
"Veritas introduced a novel approach to data replication that commanded the attention of a lot of people who thought storage-based replication was the only way to go," said Harvey Hindin, an analyst at D.H. Brown Associates Inc., of Port Chester, N.Y.
After evaluating copying data via FTP and SAN solutions, Banco Santander chose Volume Replicator because the application-based technology allowed the bank to capitalize on its existing infrastructure.
Bandwidth is a major issue for replication applications. Since Volume Replicator required minimum throughput of 512K bps and Banco Santander had a T-1 line transmitting 1.5M bps, the bank was in the clear, but not for long.
After it purchased Volume Replicator, all kinds of trafficmuch of it unrelated to financial transactions, such as e-mail and NetBIOSbegan growing and affecting replication. The traffic load became so problematic that Banco Santanders replication fell out of sync by two weeks.
Purchasing another T-1 line would have cost $60,000 per year, plus another $50,000 to $75,000 for hardware upgrades, making it unviable. Eugene Rivera, Banco Santanders technical support manager, tried to configure the banks Cisco Systems Inc. routers to give top priority to replication traffic, but that didnt work.
Banco Santander found that transmitting the same amount of data it backed up to tape40GBover the WAN, along with all the other traffic, became quite a challenge.
Help came from a sequence-caching solution offered by Peribit Networks Inc.
Veritas helped Banco Santander optimize WAN traffic using Peribits Molecular Sequence Reduction and Network Sequence Mirroring technologies. A third Peribit technology, Packet Flow Acceleration, sits at either end of the TCP connection and manages traffic flow to improve transmission.
Using Peribits tool for analyzing the WAN, Banco Santander saw that it could convert its existing T-1 and increase throughput to equal 12 T-1s.
"When we saw that, we knew thats what we need to get in here," Rivera said.
David Spark is a free-lance writer and can be contacted at firstname.lastname@example.org.
Check out eWEEK.coms Storage Center at http://storage.eweek.com for the latest news, views and analysis on enterprise and business storage hardware and software. Be sure to add our eWEEK.com storage news feed to your RSS newsreader or My Yahoo page: