Measuring Tape Storage at the Exascale

essidsolutions

It was a monumental, although not wholly unexpected next step when data storage specialist Spectra LogicOpens a new window unveiled the world’s first library to permit the handling of two exabytes of data last week.

Already a scale-tipper, big data is only getting bigger as technologies such as 5G and machine learning drive volume and demand. So while growth is inevitable, the twist in the tale comes from the integral use of an old standby: magnetic tape.

In a world of new inventions, trusty magnetic tape is rising to the challenge of archiving all that digital information and allowing access at speed, despite the fact that it has been in use almost since the beginning of the modern computer age.

The company’s line of enterprise-class libraries draw on IBM’s TS1160 tapeOpens a new window tech to house and deliver up to two quintillion bytes of data in a single library. Advances in tape materials and tape-head and cartridge constructions allow the Colorado-based company to achieve new standards in data storage and throughput.

The top-of-the-range TFinity ExaScaleOpens a new window lets enterprise users store 20 terabytes of information on a single tape cartridge and to access that data at 400 megabytes per second using applications-specific chips developed by Big Blue.

The TFinity ExaScale machine can accommodate up to 144 drives, enabling enterprise users to exceed 2EB of storage. At a compression ratio of 2.5-to-1, the transfer rate rises to 900mb/s over the fiber channel and Ethernet connections at work in most datacenters.

Tape Isn’t Going Out of Style

Magnetic tape’s central role in data storage dates back more than 60 years, thanks to favorable cost and stability. Its design consumes little or no energy when not in use and can be relied on for decades of data storage without threat of degradation.

Over the course of that long history, developments have included changing material compositions that permit ever-larger densities and embedded encryption drives that enhance security. The TS1160 features barium ferrous particles in the tape substrate to boost signal clarity.

Even as disc drives have become central to datacenter servers and personal computer operating systems, tape tech’s development has continued apace, with makers driving up both densities and the speeds with which data can be written, read and retrieved from tape libraries. Costs for tape also continue to fall to approximately one penny a gigabyte, even as the medium shifts from cheaper ferrous oxide material to more expensive composites.

Big Blue pioneered the shift away from open-reel tape – akin to reel-to-reel for enterprise-class and research computers and cassette tapes for home use – when it introduced cartridge housings in the mid-1980s.

It was also among the pioneers of the Linear Tape Open (LTO) initiative that permits greater interoperability.

Introduced in 2000 and now in its eighth iteration, the Ultrium LTO standard lets users access and store data with older cartridges. Standard LTO-8 tape carries up to 12 terabytes of raw information on a single cartridge.

Spectra Logic’s One-Size-Fits-All

In addition to IBM’s TS1160 tape format, Spectra Logic’s machines also accommodate the earlier TS1155 and TS1150 formats and Oracle’s T10000 drives in formats A though D, meaning they can be combined and deployed in a single library. They also reduce handling times with the use of 10-tape TeraPack cartridges.

The pace of the LTO development cycle has shortened to around 24 months thanks to the strides made in disc storage. For example, the Spectra Logic libraries based on the LTO-8 standard use disc-borne technology magnetoresistance tunneling to write and read data with quantum precision on thin-layer substrates.

The latest iteration features some 246,000 tracks, each separated by less than 50 nanometers, that can accommodate more than 800,000 bits of data per square inch of tape littered with barium-ferrous particles, a composite that increases the capacity of an LTO-8 cartridge to 20TB of raw and 60TB of compressed data.

The roadmap for the Ultrium standard overseen by IBM, Specta Logic and others foresees a 12th version that features upwards of 200TB of raw storage on a single cartridge, predicted to arrive by 2025.

As is common with tape’s historically cost-effective archiving, use cases for the machines that hit the market next month are envisioned in the cold storage applications common in bio-informatics and life sciences, media and entertainment, and oil and gas exploration – and, for very large data sets, the kinds used for training in machine-learning applications.

Improvements in data retrieval speeds as converged Ethernet structures replace fiber channel constructions in the cloud and on premises put tape on par with disc storage in the datacenter.

With the data deluge to come in 2020 from 5G’s expanded frequency spectrum, Spectra Logic experts expect tape’s continued development giving it a role in low-latency, high-performance computing applications, too.