Hadoop, Web 2.0 get a path to tape for cheap long-term storage
An appliance designed for Web 2.0 companies may take them back to the future with tape storage.
Tape has been around since the dawn of electronic computers and is still used for many types of long-term storage, but it hasn’t played much of a role in modern Web-based applications. When consumers go to check out old photos on social media sites, it’s likely those images will be stored on low-cost spinning disks, though Facebook is now looking to flash for this type of data.
Now tape specialist Spectra Logic is introducing a software interface and an appliance to help companies with huge amounts of less-used data shift it over to tape libraries. Tape is the cheapest way to store files for a long time that don’t need to be used frequently, said Molly Rector, Spectra’s executive vice president of product management and worldwide marketing.
Spectra Logic is jumping on what some industry analysts call a growing role for tape in some organizations with big data. Overall sales of tape systems are falling, but individual deployments are getting bigger, analysts say. Low cost and energy demands are among tape’s calling cards.
On Thursday, Spectra Logic introduced the BlackPearl Deep Storage Appliance, which uses the company’s recently announced DS3 interface. The appliance can work with special software clients for Hadoop and other big-data environments or with clients that enterprises and software vendors write themselves, according to Spectra.
Until now, Web and cloud data have had no direct interfaces to tape storage systems, Rector said. Tape libraries have special requirements, including the need to find the correct cartridge and retrieve it. “You have to know how to talk to robots,” she said. That has meant complicated middleware that Web and cloud companies don’t want to deal with, according to Rector.
DS3 and the BlackPearl provide the missing link, using REST (Representational State Transfer), an architectural approach that’s commonly used in Web and cloud environments, Rector said: “Now they can just write a normal REST application and they don’t have to know anything else beyond that.” The appliance does all the rest, including managing the data transfer process and assigning object metadata, according to Spectra. For now, the BlackPearl is designed for use only with Spectra Logic tape libraries.
Behind the appliance, a tape storage system can scale up from 15 terabytes to multiple petabytes, Spectra said. That suits the needs of enterprises in entertainment, life sciences, oil and gas exploration, government, finance and social media, as well as cloud storage providers, according to the company. Yahoo has already started using the technology to archive Hadoop data, Rector said.
When users want to retrieve a piece of data in tape storage, it will typically take about 60 seconds, though a smaller version such as a thumbnail of a photo may arrive sooner, Rector said. The big advantage of tape is its low cost, according to Spectra: A Black Pearl with a tape library ranges from US$0.09 to $0.14 per gigabyte for typical configurations, and even less for very large systems of 10 petabytes or more.
That’s cheaper than long-term storage based on hard drives or cloud storage services, and users can access the data as often as they want at no extra cost, which isn’t always the case with cloud storage, Rector said. Tape is also the most space-efficient storage technology and has a typical shelf life of 30 years, according to the company.
The BlackPearl is available for beta testing now and set to go on sale in December. The Hadoop DS3 client and the Spectra Logic Developer Program for enterprises and software vendors are available now.