Direct-Attached Storage Coming of Age in Web 2.0 World

Good old direct-attached data storage, the original digital storage model that dates back to IBM’s original spinning desk platter in the mid-1950s, doesn’t make a lot of news these days. Like a shy boy or girl on the sidelines at a junior-high sock hop, it prefers to let other newer and fancier technologies grab attention on the dance floor.

But this oh-so-basic storage form, in fact, is what’s making most of the business world go ’round. In addition to its common use in small businesses and home offices, many larger enterprises—including Web 2.0 companies—are rediscovering DAS and are starting to add it as an adjunct layer for specific mission-critical applications.

DAS is storage that is physically connected via cable or other wired connection to dedicated servers, desktops, laptops, thin clients or handheld data origination terminals.

Despite all the talk about iSCSI, Fibre Channel and 4/10GB Ethernet connectivity, industry experts estimate that DAS still comprises about 70 percent of the entire data storage market. This, of course, includes the millions of simple, USB-attached external drives individuals use in homes or in remote corporate offices, but it also includes DAS as an integral part of a growing number of enterprise storage systems.

IT analyst company IDC, which tracks various sectors in enterprise network storage each quarter, does not research DAS because of its history as primarily a consumer-focused market. IDC’s Brad Nisbet, however, told eWEEK that the company is considering adding the category to its research offerings due to the rise of DAS in the small and midsize business space.

The main reasons for this upswing? Dependability, simplicity of use, better products and falling hardware prices. All DAS requires is a direct physical connection from server to storage array; the corresponding nearness to the information source means fast performance. There’s nothing fancy about it—no hypervisor, virtual I/O or other augmentation is needed.

Entrepreneurs starting new Web 2.0-type companies are looking into DAS-based systems to anchor their growing server farms. Smaller businesses are considering housing their Tier 1 business storage on anywhere from one to four DAS machines, because scalability has been improved greatly and costs have fallen.

All in all, this is a good time to buy storage of any kind, due to the number of competitors in the sector and the ever-increasing capacities and improved software quality within the systems.

Dell and Hewlett-Packard, two of the four largest storage system providers in the world, are reporting upsurges in sales of old-fashioned DAS systems, mostly for SMBs, in the last 12 months. Neither company had definitive sales numbers to report at press time, but vice presidents at both companies told eWEEK that they have seen a noticeable bump in DAS sales over the last year.

Dual- and quad-core processors from Intel, Advanced Micro Devices and LSI in array controllers, along with higher-capacity disks provided by Seagate Technology, Hitachi, Samsung, Fujitsu and Western Digital, are making DAS storage more efficient in terms of power and performance. Stiff competition among a number of storage providers, including Sun StorageTek, IBM, Fujitsu and Hitachi Data Systems, has brought prices way down.

“Over the last year, we’ve seen our DAS sales increase noticeably, mostly for the SMB market and for branch offices,” Praveen Asthana, vice president of storage for Dell, told eWEEK.

“We’ve seen DAS growing because it’s become much more capable than it used to be. And SAS (serial-attached SCSI) technology, which really changed DAS, made it scalable,” Asthana said. “In the past, you could not scale DAS. Now, with SAS, you can attach quite a lot of storage behind the DAS unit, say 90 drives—that’s a lot of storage. When you don’t have a lot of IT staff, it makes a lot of sense to start with a DAS, attach it to the server and expand it for storage needs.”

The time to move off DAS, Asthana said, is when a business grows to need more than four servers.

“If you have more than four servers [making up an entire data centre], then DAS is not going to work for you,” Asthana said. “You’re then going to need an entry SAN [storage area network].”

Why is the cutoff at four servers?

“Because most DAS units cannot support more than that. The dual controller is there, and it can support up to four,” Asthana said. “When you need more sophisticated storage, like for document sharing, virtualisation and so on, DAS isn’t going to work for you.”

Dell is seeing two other trends in the DAS sector.

“Some of the software companies are making applications that work particularly well with DAS,” Asthana said. “The reason they are doing it is because they want to lower the cost of the hardware. They [have an incentive] to make an overall lower-cost solution for the customer.”

The other trend that Dell and HP both said they are seeing is a movement toward large clusters of DAS boxes in the Web 2.0 space that create scalable storage farms. Both vendors are competing for the expanding storage needs of social networking (MySpace, eHarmony), e-commerce (eBay, Amazon.com) and financial companies (Citigroup, Wells Fargo) that are major-league players in this space.

“New clustered DAS implementation allows for improved availability,” Asthana said. “A lot of companies are looking for that. Facebook is the kind of company we are selling these clustered DAS units to.”

Terri McClure, storage analyst for Enterprise Strategy Group, told eWEEK that enterprises like the ones noted above are sparking “renewed life” in the Web 2.0 space.

“What we’re seeing in Google GFS, Ibrix implementations—what they’re doing is layering their file systems on top of a bunch of distributed servers, making them distributed file systems,” McClure said. “And they have the capability of using DAS on the back end. What they’re really building is a grid computing solution leveraging DAS.” These are just a bunch of loosely coupled servers with commodity-based DAS, McClure said.

What kind of data is being stored in these enterprise DAS systems?

“What’s being stored is two or three copies of the data being spread across the grid infrastructure,” McClure said. “An Ibrix system, for example, that does load-balancing across it can send the data to the least-busy server to deliver that copy of the file off that server. In a big environment like Google and Yahoo have—that’s the type of implementation we’re seeing.”