First came Fibre Channel, then iSCSI. Now, for the increasingly popular idea of using a network to connect storage to servers, there's a third option called ATA over Ethernet.
Networked storage makes it easier to share storage efficiently among many servers, add storage capacity, and centralise storage systems management. But for those accustomed to the cost of storage on hard drives built into a server, the price tag can be daunting. It's a stumbling block for some to switching to technologies such as Fibre Channel, iSCSI or network-attached storage (NAS), another mainstream option.
Enter ATA over Ethernet, which uses a regular Ethernet network to link a computer to a group of ordinary, inexpensive ATA hard drives — the kind used in PCs. Coraid, a 20-employee start-up that's commercialising the approach, has found some adherents.
"It is, indeed, a lot cheaper than iSCSI," said Tim Frederick, who selected Coraid systems on two occasions for backing up data at the Colorado-based University Corporation for Atmospheric Research. "And the software integrates nicely into our Fedora Linux environment, whereas we believe we would have a great deal of complexity to deal with when it comes to iSCSI."
Some significant caveats mean that not everyone is so keen on the technology. For a start, it's a specification from Coraid, not an industry standard. Its networking abilities are limited. And its detractors include storage heavyweights such as Hewlett-Packard and Network Appliance.
With networked storage sales growing faster than storage sales overall, according to an IDC study, it's no surprise that Coraid is angling for an edge. The market for networked storage devices increased 15.4 percent, to $2.8bn (£1.5bn), from the first quarter of 2005 to the same quarter in 2006, while the total storage market rose just 6.7 percent, to $5.8bn.
How it works
ATA over Ethernet gives customers the ability to share storage and more storage capacity than can fit in a single server, but it is cheaper than the prevailing storage area network standards, Fibre Channel or iSCSI.
Fibre Channel uses special-purpose adapters to link servers and storage systems. iSCSI, which arrived later, uses regular Ethernet adapters and routes data over the network using the same technology as the Internet does: TCP/IP (Transmission Control Protocol/Internet Protocol).
And there aren't big financial risks in trying ATA over Ethernet. "The big motivating factor for using this compared to iSCSI or NAS is how inexpensive it is," said one Coraid customer, who asked not to be named. "If it doesn't work, we're down $3,000, but all the disks we can use in other places."
Used for networked storage, the technology is well-suited to tasks such as keeping medical records, online photos or surveillance videos, said Jim Kemp, chief executive at Coraid.
The company's most popular product is the SR1520, a 5.25-inch-tall system that accommodates 15 drives and uses RAID (redundant array of inexpensive disks) technology for increased data reliability. The system costs $3,995, not including disks.
Like iSCSI, ATA over Ethernet uses conventional networking hardware. But in contrast, it doesn't use TCP/IP. That's a mixed blessing: While systems using ATA over Ethernet aren't burdened by processing TCP/IP networking information, they also can't take advantage of the technology's network-routing abilities. That limits ATA over Ethernet to being used on local networks.
That's just fine with UCAR's Frederick. He installed a separate network on servers that connect to the Coraid backup device.
"We keep everything within a rack and dedicate the second Ethernet on our machines for ATA over Ethernet. We also dedicate a switch. So, no, we haven't had concerns over the local area network limits," he said. The second network provides some more security, he added.
It's not a miracle cure, however. The biggest disadvantage of ATA over Ethernet is that it's slower than direct-attached storage using SCSI, he said. And he uses Coraid only for backup rather than for primary storage.
Detractors
HP, which sells both servers and storage systems, isn't a fan of ATA over Ethernet. In effect, it functions more as a new method for directly attaching storage to a server, said Dwight Barron, chief technologist for HP BladeSystem. Customers wanting that technology are likely to favour new storage systems that pack several Serial ATA (SATA) or Serial Attached SCSI (SAS) drives into a chassis, he said.
"Had SAS and SATA had not stepped up to the role they have, (ATA over Ethernet) may have had some play for some audience, but we're betting SAS and SATA will fulfill that need," Barron said.
And using Ethernet but not higher-level TCP/IP network control technology is a serious limitation, Barron added. "What are you really going to do with a protocol over Ethernet that's not routable?" And not using TCP means that there's no network traffic congestion management, he added.
Another hurdle for ATA over Ethernet is formal standardisation by several companies. "It's not an industry standard and appears not to be on the road map for any industry standards. That's the obstacle to broad adoption," said Dave Dale, NetApp's industry evangelist and chairman of the Storage Networking Industry Association's IP Storage Forum.
Standardisation will happen later, though. "We will work with standards bodies as we grow the company and have the resources to spend in that area," Kemp said.
Kemp is convinced that despite the industry skeptics, high prices for the alternatives will keep customers coming. "They're pretty shocked by the cost of Fibre Channel and iSCSI. They discover ATA over Ethernet, and it's a breath of fresh air," he said.
