IBM Gets Tape Ready for ‘Big Data’
May 16, 2011 Alex Woodie
The leading purveyor of Big Iron last week unveiled a slew of new tape products designed to handle “big data.” Among the enhancements IBM introduced is a new mechanical shuttle system that links up to 16 TS3500 tape libraries, providing almost an exabyte of native storage capacity in a single archive. The vendor also announced the fourth generation of its propriety 3592 series of enterprise tape technology, and made a slew of enhancements to disk-based storage offerings.
“Extreme scalability” is how IBM describes the storage capacity enabled by the new TS3500 Tape Library Connector Model SC1, a mechanical shuttle system that enables tape cartridges to be moved from their “home” libraries to up to 15 other connected TS3500 Model S24 or S54 libraries. A fully populated TS3500 string could enable a customer to house 300,000 LTO tape cartridges in a single library system. That equates to 450 PB of native storage capacity when using the latest LTO5 gear, which has a native capacity of 1.5 TB per cartridge.
The new TS3500 shuttle system, which starts at $20,000, gets even more extreme when equipped with the latest TS1140 tape drive technology, the fourth generation of IBM’s 3592 Enterprise Tape Drive that IBM also updated last week. With 4 TB crammed into each 3592 Advanced Data (Type C) cartridge (also unveiled last week), the total capacity of a fully outfitted TS3500 string doubles to 900 PB, or nearly an exabyte of uncompressed capacity. And when 3:1 compression is turned on, the capacity jumps to 2.7 exabytes. Take a look at this beast:
That’s a lot of data. (One might even call it “big” data, if one were so inclined). To put that in perspective, 2.7 exabytes is about three times all the mobile data generated in the United States last year, according to IBM. It’s also nearly three times as much as Oracle can cram into its largest tape library, the StorageTek SL8500 Modular Library System. That Oracle library can hold 500 PB of uncompressed data (or 1 EB with 2:1 compression) when fully outfitted with 100,000 cartridges written with Oracle’s latest tape drive, the StorageTek T10000C.
With 5 TB per cartridge of native capacity, the StorageTek T10000C technology is the densest in the industry. However, IBM’s TS1140 drive and the 3592 cartridge offer other advantages. For starters, the setup is faster, at 250 MB/sec compared to the 240 MB/sec of the StorageTek product (360 MB/sec with compression). The max speed of the TS1140 drive with compression is 650 MB/sec.
IBM also claims that its new TS1140, which has a list price of $42,995, uses 64 percent less energy than the Oracle T10000C tape drive. The TS1140 operates at a maximum power threshold of 51 watts, compared to 67 watts for the T10000C drive, according to IBM.
Fierce competition between IBM and StorageTek, which was founded in the 1970s by former IBM engineers, is nothing new. IBM has battled StorageTek in the tape library with the precursors to the TS1140, including the 3480 line, which launched in the 1980s, and the 3590 products, which launched in the mid-1990s under the “Magstar” brand. However, the 3592 is not compatible with the 3590.
Only tape can provide the price/performance demanded by the fast pace of world data growth, says IBM, which says it owns 40 percent of the world’s market for tape products. Big “Data” Blue cites an August 2010 statistic from IDC that says data is growing at a compound annual growth rate of 49.8 percent through 2014. This data is being generated and stored on the Internet through mobile devices, social networks, cloud computing, and “armies of sensors,” says IBM.
Tape is well positioned for this growth. In fact, according to an Enterprise Strategy Group statistic cited by IBM, the amount of data stored on tape archives is expected to grow by a factor of six between 2010 and 2015. In particular, IBM sees tape gaining additional traction in the media and entertainment and healthcare, which require lots of cheap storage for their huge data archives.
IBM made several other storage related enhancements last week, but few were pertinent to the IBM i system or IBM i customers. This includes the new Linear Tape File System (LTFS) for IBM’s LTO5 gear. The LTFS was one of the new features in the LTO5 specification when it was launched about 16 months ago. The technology lets metadata run piggyback on the actual data, thereby giving users more freedom in how the data is used. However, it only works with Unix, Linux, and Windows. IBM also announced new data replication features in its System Storage TS7650 ProtecTIER data de-duplication appliance.
IBM also announced support for 10 Gigabit Ethernet connectivity in its System Storage SAN Volume Controller (SVC) Storage Engine, a modified System x server that is used to run the SVC software that controls IBM disk arrays (SANs) that can connect to the IBM i server, such as the DS4000 and DS8000, as well as competing SANs from EMC, Hitachi Data Systems, and Hewlett-Packard.
Big Blue’s Scale Out Network Attached Storage (SONAS) offering was also enhanced. IBM developed the clustered file system to provide a high performance, inexpensive, and modular storage environment for companies with variable storage needs, such as cloud and SaaS application providers. IBM says it has doubled the throughput of its SONAS array with version 1.2 to 403,000 IOPS, which it says is a world record. SONAS-based arrays can scale up to 14 PB of storage across 7,200 SAS and nearline SAS drives, and fetches data using CIFS, NFS, FTP, HTTP, and, now, NDMP protocols.
Also announced were enhancements to the IBM Information Archive for Email, Files and eDiscovery; and enhancements to the TS7700 Virtual Tape Library family. The TS7700 VTL is based on Power7 processors, and is designed to replace tape drives in System z mainframe environments. Version 2 doubles the number of virtual tape cartridges it contains to 2 million.