Where is tape heading and does it have a future?
Currently, more than 500 exabytes of data are already stored on tape, and tape is predicted to host zettabytes of information within a couple of years.
A lot of that tape storage is going to be hidden in the cloud. While cloud vendors loudly promote their use of flash and fast disk to provide premium service, their dirty little secret is that they quietly host a lot of data on tape. Those that don’t will be forced to soon, said Fred Moore, an analyst with Horison Information Strategies.
“If you don’t have tape in the cloud, your cloud is going to burst,” said Moore.
He pointed to various TCO studies that indicate tape is much less expensive and more reliable than disk for large volumes of data. But it is what lies ahead that excites him.
A common gripe about tape is that it is sequential. Therefore, you have to feed the entire tape past the heads to find the file you need in some cases. Disk, on the other hand, is limited only by the disk having to be moved to the right place on the circumference and the actuator moving across the disk, so it is much faster in terms of access. But a survey of tape patents filed this year revealed work being done to address this shortcoming. Someone has figured out a way to bring tape access closer to that of disk.
“Partitioning of tape breaks down the sequentiality barrier and a patent has been filed for this,” said Moore. “Another patent has been filed about a method of writing multiple files simultaneously to tape.”
Yet another patent was filed for random access to tape, and another for a solid state media cartridge which combines flash and tape in one cartridge. If all this comes to fruition, tape will be an even stronger platform for archiving.
“Modern tape has the longest media life of greater than 30 years,” said Moore. “Oxidation is the biggest reason for deterioration and Barium Ferrite (BaFe) tape is already oxidized, so it lasts a long time.”
According to testing conducted by IBM, the areal recording density of BaFe tape is 123 Gb per square inch. This is 61 times the density of current LTO-6 tape and offers the potential of cartridges reaching 220 TB.
“Barium Ferrite is available for LTO-6 and will be available for LTO-7, which will be on the market soon,” said Peter Faulhaber, president and CEO of FujiFilm Recording Media.
He indicated the vast progress of tape in the last few years in terms of raw capacity. Future LTO-7 cartridge will eventually hold the same capacity as 6 pallets of DLT4 cartridges. In terms of cost, you are looking at $300 for that LTO-7 cartridge versus $330,000 for all those DLT4s. LTO-7 will initially have 6 TB native capacity (15 TB compressed) with a 300 MB/sec transfer rate. It will be out by the end of the year, said Faulhaber.
While tape had 11 percent annual capacity shipment growth from 1951 to 2000, it is now at a rate of 47 percent per year. Back in 2000, one LTO-1 cost $1250 per TB and now LTO-6 is $14 per TB, said Faulhaber. The industry roadmap shows that this magnitude of gain is expected to continue.
“Tape has far more potential available for further density achievement whereas disk is struggling on where to go next,” said Mark Lantz, principle research staff member at IBM. “Disk is more arealy dense by a considerable margin, but because tape has much more media to write on (1000 meters of tape per cartridge), it has more potential for density shrinkage.”
IBM has developed various technologies to improve the speed, density and accuracy of tape. Its TS1150 enterprise tape line has an astounding hard error rate of 275,000 hours. The company has also released IBM Spectrum Storage for archiving.
“Big data is good for tape as it is six times cheaper when all costs of media, hardware and power are taken into account,” said Ed Childers, senior technical member at IBM.
With IDC predicting 44 ZB of data in existence by 2020 and some predictions going as high as 250 ZB by 2030, Childers expects plenty of growth potential for tape.