Jewels among storage standards
- By Alan Radding, John Moore
- Aug 15, 2005
When storage was still a back-burner issue for most information technology managers, options were scarce for standards-based storage gear. You had the prevailing product type from Vendor A, B or C. No one who was crazy enough to try such a thing expected the products to work together.
Storage standards have become more plentiful and meaningful, making products more compatible and enabling different approaches to solving common data storage and management problems.
However, be warned: Even when dealing with standards-compliant gear, managers should stick with products that have been certified for interoperability with the other tools they want to use.
Federal Computer Week asked leading storage industry analysts about some of the standards that will be most important as the industry moves ahead. In the second part of this two-part article, we look at standards for storage networking, data protection and device management.
Internet SCSI (iSCSI) is rapidly gaining acceptance as an alternative to Fibre Channel for building storage-area networks (SANs). The iSCSI protocol sends standard SCSI storage commands over general-purpose TCP/IP data networks.
"We believe iSCSI is certainly ready for prime time," said Dianne McAdam, a senior analyst and partner at the Data Mobility Group. "It has even gained acceptance at the largest storage vendors, which now offer iSCSI products."
ISCSI was originally targeted at small and midsize organizations or large organizations' remote offices. Those setups need the any-server-to-any-storage benefits of networked storage, yet they lack the IT skills and resources required to deploy Fibre Channel SANs.
Using iSCSI, organizations can network their storage to create a SAN using their existing Ethernet IP networks. They don't have to learn a new networking protocol in Fibre Channel or invest in Fibre Channel components, which can be costly. All they need are inexpensive iSCSI Network Interface Cards for the servers and a new iSCSI storage array. Or they can do an iSCSI upgrade of their systems.
"We're seeing a lot of iSCSI installation in government," said Greg Schulz, senior analyst at the Evaluator Group. "It is cost-effective, and you no longer need special components."
Now iSCSI has become popular for server/
storage consolidation even at large organizations, Schulz said. The organization can connect servers that previously had their own direct-attached storage to shared storage on an iSCSI SAN. And the storage can be administered centrally instead of managed at each individual server, which lowers administrative overhead, he said.
Fibre Channel over IP
Fibre Channel over IP (FCIP) lets organizations bridge Fibre Channel SANs over IP networks. The protocol finds support among such storage switch vendors such as Brocade Communications Systems, Cisco Systems and McData.
The FCIP approach encapsulates Fibre Channel frames the fundamental units of Fibre Channel data transfer in TCP/IP packets. FCIP is referred to as a tunneling protocol. It aims to overcome the distance limits and expense of using Fibre Channel to unite SAN islands.
"Instead of incurring the cost of implementing a Fibre Channel infrastructure for joining SAN islands over wide areas, FCIP offers a cost-effective alternative," said Todd Kofchur, a consultant at Sanz, a storage integrator.
Remote backup and recovery rank among the main applications for FCIP, industry watchers say. But Prasad Pammidimukkala, vice chairman of the IP Storage Forum at the Storage Networking Industry Association (SNIA), said FCIP comes into play "any time you want toâ€¦extend a storage network application over distance over readily available IP networks."
Switch vendors provide FCIP capabilities through a stand-alone device, a specialized hardware component or blade that goes into a switch, Pammidimukkala said.
Kofchur said vendors "are beginning to offer integrated FCIP capabilities in their Fibre Channel switches that not only eliminate additional [hardware] purchases, but also provide
performance-enhancing technology such as compression."
Rick Hoffman, storage section manager for technical services in Michigan's IT department, said his organization isn't considering FCIP. Dark fiber and the dense wave division multiplexing equipment that uses the fiber give the state more than enough capacity for off-site data mirroring, he said. He added that the state's disaster recovery site is about 10 miles from its primary EMC storage arrays.
Storage Management Initiative Specification
The Storage Management Initiative Specification (SMI-S) is intended to standardize the management of vendors' storage devices. It provides a way for storage devices to communicate with storage management software, allowing the software to not only discover and monitor equipment, but also control it. SMI-S is based on the Common Information Model (CIM) but will be developed much further than that standard.
At the most basic level, SMI-S can capture basic information from a device, similar to the role that the Simple Network Management Protocol plays in network management systems.
Most large storage vendors promise to support SMI-S, but the follow-through has not been great. "SMI-S clearly needs more vendor and product support," Schulz said.
To make their products more useful, vendors can add extensions to SMI-S that enable more functions than the standard's basic specification allows. This provision could benefit customers and encourage more vendors to support the standard, because it allows them to differentiate their products with special features.
Common RAID Disk Data Format
The Common Redundant Array of Independent Disks (RAID) Disk Data Format (DDF) specification provides a common format to allow basic interoperability among RAID vendors' products, Kofchur said.
"Currently, RAID formats are vendor-specific and do not allow for data-in-place migrations of disks that are controlled by one vendor's RAID controller to another's," he said.
An organization that seeks to replace a RAID subsystem has to use the same type of subsystem for the data to be read, said Bill Dawkins, chairman of SNIA's Common RAID DDF Technical Working Group.
Use of RAID DDF would make it easier to replace a RAID card in the event of a failure or upgrade a server to the latest generation of RAID technology.
Today, customers incur the administrative overhead and costs of transferring data from an existing system to tape and then migrating the data to a new RAID platform. n
Radding is a freelance journalist based in Newton, Mass. He can be reached at [email protected]