Where ATA arrays can save you money

ATA arrays can save you money if you use them where their limitations don't have a deleterious impact.

This Content Component encountered an errorThis Content Component encountered an error
This article can also be found in the Premium Editorial Download: Storage magazine: Tame the e-mail beast: Valuable e-mail management strategies:

In the enterprise storage array market, SCSI has traditionally been the only game in town. IT managers rarely think twice about rubber-stamping orders for disk arrays using high-end SCSI hardware, especially for use in data centers and other mission-critical environments.

Today, however, SCSI may be losing its iron grip on the enterprise RAID market. A growing number of storage vendors have introduced ATA-based RAID products that offer solid - and rapidly improving - performance and reliability, combined with management tools suitable for enterprise applications. Many of these products are designed primarily for disk-to-disk backup applications.

Analysts, vendors and end users generally agree that SCSI-based RAID is still the best choice for volatile, high-intensity applications such as databases.

Yet many analysts also predict that ATA technology, as well as new products based on the next-generation Serial ATA standard will eventually challenge SCSI even in the high-end enterprise RAID market. "In the past," says Tim Klein, CEO of ExaDrive Networks, Sunnyvale, CA, "we fought the perception that ATA was all junk and that a company's data was too important to be put onto ATA. That has settled down - people understand that ATA is real, it's reliable and you can count on it."

How low can it go?
The price gap between SCSI and ATA drives is wide. As the cost of 1TB of raw ATA-based disk storage falls in some cases to less than $1.25 per GB, the gap has widened into a canyon. Many retailers, for example, now offer brand-name 80GB 7,200 rpm ATA/100 drives for less than $100; by comparison, a typical 73.4GB 10,000 rpm Ultra160 SCSI drive can cost over $350, or more than four times as much per GB.

At the same time, 7,200 rpm ATA drives with capacities up to 200GB are now widely available for less than $2 per GB. Only a handful of Ultra160 SCSI drives exceed 80GB, by comparison, and those that do carry a whopping price tag - typically $6 to $7 per GB. "The ability to get a half terabyte of storage with just four ATA drives is incredible if you think about it," says David Vaughn, a product manager in IBM's storage networking division. "We now have the ability to provide an enormous amount of storage at a very low cost."

In addition, performance and reliability differences between SCSI and ATA have narrowed, and according to industry analysts, these differences become even less significant in array-based configurations. "With more than five drives in an array, you saturate the bus on just about any system out there, and you can't take advantage of the faster data transfer provided by higher [SCSI drive] spin rates," says Bob Ripley, vice president of marketing for Consensys. "If a SCSI drive spins out data three times faster per drive on a large array, all you're doing is generating heat."

Some industry experts also challenge the assumption that ATA disks aren't reliable enough for enterprise use. "I haven't heard any derogatory statements made about ATA reliability, and in many cases I hear just the opposite - ATA drives can be more reliable than some SCSI and Fibre Channel [FC] drives," says Peter Gerr, research analyst for the Enterprise Storage Group, Milford, MA. And with MTBF rates on ATA drives of 30 years or more, disk failures in RAID systems are a real problem only for the most critical enterprise applications.

Gerr also stated that so far, SCSI vendors haven't moved much to cut prices on their products, although that could change. "There's the potential for downward price pressure on SCSI and Fibre Channel products, but it's still too early to tell," he says.

New vendors
A growing number of vendors now offer ATA-based RAID products. Many of these firms focus on the small business and low-end server markets with RAID products supporting a small number of direct-attached ATA drives. Others have focused on "RAID on a chip" ASIC solutions designed to move disk array support onto the system motherboard.

But a handful of companies - including 3ware, Consensys, and ExaDrive - have set their sights on the enterprise RAID market. These firms have developed ATA-based direct- or network-attached RAID products designed to match or even exceed the performance of SCSI- or Fibre-based arrays. Consensys' Raidzone product line, for example, uses a proprietary backplane and controllers that the company claims will equal or exceed the data throughput of systems using Ultra SCSI bus. "We don't give up anything to the SCSI and Fibre Channel arrays except the price," Ripley says. Other ATA-based RAID products, such as ExaDrive Networks' Diamond series, use a hybrid approach: They combine ATA disks with multiple SCSI or FC host interfaces and a proprietary controller.

According to the manufacturers, these ATA-based RAID products offer many of the same benefits as SCSI-based arrays. They typically support all standard RAID levels, including RAID 5, and they include hot-swap and hot-spare capabilities with automated failure detection, failover and data remirroring. They also provide policy-based management tools, remote monitoring and advanced diagnostics such as temperature, voltage and electrical current monitoring. Although ATA-based RAID vendors don't claim to match the largest enterprise RAID installations, some customers have built ATA-based network-attached storage systems using dozens of disks with 10TB.

ExaDrive Networks' Klein also noted that ExaDrive certifies the interoperability of its arrays with a variety of third-party products, including FC switches, storage virtualization, SAN management tools, and operating systems. "We approach the market differently than the big turnkey vendors. We say if you're looking for a best-of-breed approach, that's where we come in," he says. "The hardware savings give customers a margin to figure out where to get the best software solutions from high-end VARs and integrators. That's why we put a big emphasis on interoperability, especially on the Fibre Channel side for [storage area network] SAN implementations."

Serial ATA and iSCSI
Serial ATA (SATA), introduced in August 2001, offers several significant areas of improvement over ATA. The most noticeable improvement is data bandwidth: Serial ATA, which combines a new hardware interface with the existing ATA command protocol, supports data transfer rates up to 150MB/s - the next version of SATA will increase that to 300MB/s. SATA uses much smaller cables with a reduced pin count and a maximum length of one meter, allowing easier cable routing, improved system airflow and greater scalability. SATA also uses a lower operating voltage, as well as command queuing and other performance enhancements.

Some industry analysts believe that SATA could further close the already narrowing gap between ATA and SCSI disk drives. "I'd say that in the top 25% or 30% of the enterprise market, Fibre Channel drives will be the better choice," says Gartner's Roger Cox. "But from a technology point of view, the other 70% of the enterprise market could easily be satisfied by Serial ATA technology."

In order for SATA to make its mark on the enterprise market, however, major storage vendors will have to introduce products using the new standard and market them aggressively - something they have been slow to do. According to Cox, increasing competitive pressures will quickly force vendors to get serious about their SATA product plans. "I think someone like EMC coming out with a high-end network-attached storage product using Serial ATA would seize a tremendous breakout opportunity in this market," he says.

Another emerging standard, Internet SCSI (iSCSI), could further expand the reach of ATA-based storage in the enterprise. iSCSI, which specifies an IP-based networking standard for linking data storage facilities, could eventually challenge FC as a cost-effective solution for building storage area networks. "Serial ATA-based arrays, used in conjunction with iSCSI routers, storage switches and other products, could be a perfect match, especially in the midrange market," says Peter Gerr, research analyst for the Enterprise Storage Group.

ATA finds its niche - and more
So far, ATA-based RAID has received the warmest welcome in certain vertical markets. "Our major users include government and university research laboratories, oil and gas exploration, video surveillance, data storage and other customers who produce very large amounts of data," says Consensys' Ripley. "In many cases, there's no other affordable way to store that data without using tape or optical storage, and that means giving up immediate access to the data."

According to Scott Studham, group leader in the Pacific Northwest National Laboratory's molecular science computing facility in Richland, WA, the cost-performance relationship between ATA and SCSI was a decisive issue in the group's decision to invest heavily in ATA-based disk arrays. Studham says the laboratory recently migrated its supercomputer data archives from a 20TB tape-based system to an ATA-based storage array. In addition, he says the laboratory will use a 200TB ATA-based RAID array as primary storage for its new supercomputer, which he says will be the world's largest and fastest Linux-based system.

"A lot of people asked why we weren't using SCSI [RAID technology], but once I presented the facts, they all understood," Studham says, referring to the conclusions he reached about the performance and reliability of ATA-based disks vs. more expensive SCSI hardware. "In this environment, that 4% to 5% performance penalty for using ATA instead of SCSI doesn't outweigh the 200% to 300% price difference."

In the mainstream enterprise market, ATA-based RAID has also emerged as an alternative to tape-based backup systems and to nearline systems such as CD jukeboxes. As disk prices fall and ATA-based RAID products become more sophisticated, these products have created a new and rapidly-growing niche in the nearline storage market.

According to Jamie Riis, CIO of BayView Financial, a mortgage bank based in Miami, FL, moving to ATA-based nearline storage yielded significant benefits. BayView currently uses a Network Appliance NearStore R100 system to store backup data snapshots before moving them over to tape-based storage, allowing the company to provide almost instant access to recent backups. "It could take up to 20 hours to get this data from the tape system," Riis says. "Now if someone loses a file, we can go back at least to where they were at the last snapshot and restore the data in a few minutes."

In addition, Riis says BayView also plans to migrate its document image archives from an optical jukebox to the R100. "We'll go from a system where it takes 30 to 40 seconds to retrieve an image and we have to throttle user access, to an environment where hundreds of users are getting submillisecond access," he says.

According to Chris Bennett, director of product marketing for Network Appliance, Sunnyvale, CA, Riis' comments reflect an important trend for ATA-based enterprise storage: "ATA drive technology is enabling the industry to store information online that previously was not available," he says. "When you're comparing the amount of data now stored online to the amount stored on tapes stacked in cardboard boxes someplace, you're talking about orders of magnitude of difference."

Major vendors respond
Some top-tier vendors - such as IBM, Network Appliance, and EMC - have released ATA-based RAID products designed for nearline storage, disk-to-disk backup systems, or even as mainline storage for mid-market enterprise customers (see "Major vendors introduce ATA RAID products.")

Unlike their smaller competitors, these firms are more conservative about the short-term prospects for ATA-based RAID. "We don't yet see customers who are willing to put mission-critical data on ATA drives," Vaughn says. "You still get better performance with SCSI technology, and that performance advantage is clear. The ATA platform is important, but we're using it to target the midmarket and distributed environments right now."

Nevertheless, according to Bennett, Network Appliance was surprised by the response to its ATA-based NearStore product line, which began shipping in January 2002. "We've shifted significant resources onto NearStore because of the exceptional response," he says. "But we also want to position it where it's appropriate. It's not just a performance issue, it's an availability issue - the [ATA] drives are slower and more prone to failure, and you run a higher risk of degrading critical operations in an ATA environment."

ATA in the data center?
Although ATA-based RAID has gained ground in niche markets and as an alternative data backup solution, the question remains: Will ATA present a serious challenge SCSI in the data center?

Gerr says he believes many enterprise users are prepared to take their cues from the major vendors with whom they have long-term relationships. "It takes more market leaders in the industry to adopt ATA as an important part of their technology roadmaps to help users feel more comfortable about ATA as a viable alternative. It's a matter of psychology, of marketing, of testing and experience," he says.

According to Vaughn, however, resistance to ATA in mainstream enterprise RAID applications runs deep. "Enterprise customers are absolutely willing to consider ATA in distributed environments, but not in the data center," he says. "If you talk to the data center guys, they aren't willing to risk their careers just to save money by using ATA instead of SCSI."

Major vendors introduce ATA RAID products
In recent months, IBM, Network Appliance, and EMC have all introduced ATA-based RAID products. These companies' ATA products aren't designed for mainstream enterprise use - rather, they're intended for niche applications or for midmarket customers that traditionally couldn't afford enterprise-class RAID products. Nevertheless, these products represent an interesting development in the high-end storage market, and they could pave the way for additional ATA-based RAID solutions.

Network Appliance and EMC have introduced ATA-based RAID products designed primarily for nearline storage applications. The NetApp NearStore 100, with 12TB of ATA-based RAID storage per appliance, consolidates data from multiple filers, creating an accessible backup source for recent data and providing a platform for transferring backup data to offline storage. EMC, by comparison, positions its ATA-based Centera system as a nearline archive for static content: data such as medical imaging, rich media content and other material that users require occasional access to without the inconvenience of accessing offline storage systems. EMC doesn't describe Centera as a RAID product, although its functionality is quite similar.

IBM, on the other hand, has moved closer to providing a true ATA-based RAID solution with its TotalStorage NAS 100 system. The NAS 100, which includes four 120GB ATA drives per unit and provides RAID 0, 1 and 5 support, is designed primarily for customers with multiple branch or remote locations, or as an entry-level RAID solution for small- or medium-sized business customers. In either case, IBM expects the NAS 100 to provide a cost-effective solution, allowing customers to adopt RAID technology in situations where they might otherwise use nonredundant storage.

In fact, some IT managers are wary of ATA technology precisely because its cost appears too good to be true. According to Matthew Stock, director of information technology for the computer science and engineering department at State University of New York at Buffalo, his organization is not yet willing to move ATA-based RAID systems into high-performance environments in spite of its lower cost. "We've worked with the ATA systems, and it's a really good value for the dollar if you need a lot of storage and aren't too concerned about performance," he says. "But part of me thinks, if I'm getting 2TB of ATA storage for what it would cost for 300GB in a NetApp or EMC system, I've got to be losing something, whether it's the value-added software, or the performance, or the reliability."

In other cases, however, IT executives have decided the marginal performance benefits of SCSI simply aren't worth the price. Auctionwatch.com, San Bruno, CA, a major provider of online sales management applications and services, has almost completed a company-wide migration from SCSI to ATA-based storage arrays. According to CTO Greg McNutt, although the move was initially cost-driven, the company gets virtually the same level of performance and reliability it had with SCSI storage - including on critical database and application servers. "On our systems, the performance has been nearly equivalent," McNutt says. "Even on our databases, where we initially had some concern about performance issues, right now the evidence is that the ATA drives aren't any better or worse [than SCSI]."

Mike Wentz, director of product marketing for 3ware, in Mountain View, CA, says he believes resistance to ATA will continue to wane as managers gain more exposure and learn to accept the technology. "The price point of ATA is so attractive that companies can use it for data that didn't have the benefit of RAID protection in the past," he says. "This gets ATA into the data center through a bottom-up approach ... as CIOs and IT managers get comfortable with ATA technology on the low end, it will begin to migrate up the hierarchy.

Roger Cox, a chief analyst in Gartner Dataquest's Storage Research organization, observed that other storage technologies traveled a similar path into the data center. "You have to fight customer perceptions. They'll say, 'ATA technology is junk. I don't want that, I want SCSI and Fibre Channel.' But we're going through the same experience with ATA drives that we went through 15 years ago, when EMC and other companies introduced drives based on small form factors at a time when they weren't considered adequate for enterprise use," he says. (See "Serial ATA and iSCSI.")

Accepting the inevitable
Do top-tier storage vendors have an interest in slowing the spread of ATA-based RAID products? The thought has occurred to some competitors who point out the unusually high margins associated with SCSI storage hardware, which in many cases costs two to three times more per GB of storage than ATA disks.

"I think they're being dragged kicking and screaming into the ATA market," Ripley says of the major vendors. "You see a huge price gap between their ATA and SCSI products. They're trying to come up with creative marketing reasons why one is different from the other, but it's all pretty transparent. To be fair, it's a huge problem: They're used to selling at very high margins, and the ATA business is based on low margins and low profits."

According to Studham, some vendors were reluctant to work with his organization on ATA-based RAID solutions. "I think it's just a marketing decision on their part," he says. "If you have a market selling high-performance SANs for $50,000 per TB you'd be shooting yourself in the foot to come out with a product that had the same performance characteristics at $5,000 per TB."

Cox agrees that some vendors view ATA-based RAID as a threat rather than as a new business opportunity. "To me, a lot of this isn't about technology. It's about protecting revenue and protecting business," he says. "Companies like IBM, Dell, and HP are reluctant to sell ATA because they don't make as much money on it, but eventually the competition will force them to change."

According to Bennett, however, Network Appliance has already embraced ATA as a rapidly growing part of the enterprise storage market. "These lesser technologies generally start with poor performance and low reliability, but their main benefit is that they're cheap and just good enough to work in niche markets. But as they get better over time, they get traction and move up-market," he says. "It's a classic example of disruptive innovation - the incumbents fight it to protect their installed base, and the upstarts embrace the technology and drive it forward. But you either deal with reality or you get run over, and our plans are to embrace this stuff actively."

This was first published in October 2002

Dig deeper on Disk arrays

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

-ADS BY GOOGLE

SearchSolidStateStorage

SearchVirtualStorage

SearchCloudStorage

SearchDisasterRecovery

SearchDataBackup

Close