Two groups of vendors have set up separate transcontinental and global storage networks based on emerging IP-based data-transport standards. But the storage-over-IP technologies still aren’t expected to be robust enough for enterprise uses for at least another year, say analysts.
IP-based storage has several potential advantages for users. For example, it would use the Internet’s basic communications protocol instead of more expensive Fibre Channel connectivity technology.
In addition, Fibre Channel links are currently limited to distances of about 100 km because the data packets break down quickly. That’s a problem that forces IT managers to back up data to storage devices in nearby data centers.
In one recent demonstration, IBM Corp., Dell Computer Corp., Intel Corp. and five other vendors used off-the-shelf storage and switching equipment to transfer block-level data over IP between sites on the East and West coasts (see diagram, below).
Database records and other information were transferred over a 10G bit/sec. fiber-optic backbone at speeds of up to 2.5G bit/sec., the vendors said.
The switched storage-area network (SAN) sent about 1TB of data per hour between systems in Sunnyvale, Calif., and Newark, N.J., they added. The network was based on emerging standards known as iSCSI, Fibre Channel over TCP/IP (FCIP) and iFCP (see box).
Meanwhile, Compaq Computer Corp. and two other vendors said they used FCIP to manage and replicate data among Fibre Channel SANs in Colorado Springs, Australia and the Netherlands.
Steve Duplessie, an analyst at Enterprise Storage Group Inc. in Milford, Mass., said the IP-based storage tests show that “the stuff works,” even though the proposed standards haven’t been finalized.
Standards are expected to emerge from the Internet Engineering Task Force by year’s end. And vendors such as IBM and Cisco Systems Inc. are already hawking disk arrays, switches and software that can be used on intranets and other IP-based networks. But Duplessie and other analysts said it will likely be late next year before IP-based technology is ready to support enterprise-class storage applications.
Brian Haymore, a senior systems engineer at the University of Utah in Salt Lake City, said he sees iSCSI as the future of data storage. Haymore is testing eight storage arrays that can use IP to map block-level SCSI commands and data over an Ethernet network. With IP, he said, the mapping can be done on the school’s existing extranet and be managed by IT staffers who are already familiar with TCP/IP.
Mike Anderson, vice-president of information services at The Home Depot Inc. in Atlanta, said IP networks hold promise for storage-related uses such as real-time data replication in backup and disaster-recovery applications.
But Anderson added that he wants to see increased product maturity and widespread user adoption before he considers iSCSI or other storage-over-IP technologies.