128TB tape cartridges key to kilometer-size telescope
IBM has already demonstrated it can shrink areal tape density by a factor of 10
Astronomers building a kilometer-sized radio telescope are depending on 60-year-old magnetic tape technology to store the 1 million GB of data per day they plan to generate.
The Square Kilometre Array (SKA) is a radio telescope being built by the Netherlands Institute for Radio Astronomy (ASTRON) in Australia and South Africa. The telescope arrays will consist of 3,000 dishes, each 15 meters in diameter and is expected to cost about $2 billion. Astronomers and engineers from more than 70 institutes in 20 countries are designing the SKA, which is expected to be 50 times more sensitive, and will survey the sky 10,000 times faster, than today's radio telescopes.
The dishes will look for new galaxies, dark matter and the origins of the universe.
An artist's rendition of the SKA telescope (Source: ASTRON)
The radio telescope will be so sensitive that it will be able to detect an airport-style radar on a planet 50 light years away.
In one day, the telescope's dishes will generate 10 times the network traffic produced at the same time on the global Internet. They will feed about 10 petabits of data (1 billion gigabits) per second into a central computer that will have the processing power of about 100 million of today's PCs.
The SKA supercomputer will perform 10 18 operations per second, equivalent to the number of stars in 3 million Milky Way galaxies.
ASTRON has partnered with IBM which, under a five-year contract, is developing the exascale computer system for processing the deep-space data. IBM is also responsible for the data storage technology, and for that, it is reaching back to magnetic tape, but this isn't granddad's technology.
To offer some idea of what IBM is attempting to achieve, imagine a cartridge with 1,000 meters of half-inch wide magnetic tape. Then imagine a tape drive trying to position a read-write head on tracks within a 10 nanometer-wide area (a nanometer is one billionth of a meter), and it's doing that while the tape is moving at a velocity of 7 meters per second.
"We were able to demonstrate [read-write head] accuracies in the ballpark of 29 nanometers, and we have some experimental evidence showing we can get to 15 nanometers," IBM Fellow Evangelos Eleftheriou said, "so there's a little bit more work we have to do to get to 10 nanometers."
Even at 29-nanometer read-write head accuracy, IBM has proved it can achieve 10 times more accuracy than what's available in corporate data centers today, Eleftheriou said.
The SKA will need to store 300 to 1,500 petabytes (1.5 exabytes) of data per year, generating enough raw data to fill 15 million 64GB iPods every day. It wouldn't be cost effective to store that amount of data on hard disk drives, said Eleftheriou, who heads the Storage Technologies Group of IBM's Zurich Research Laboratory. Today, tape costs about $25 per terabyte of capacity.
"Eventually, you store [the data] in files that have to be accessible by scientists around the world. It's similar to what's happening in CERN, but even bigger," Eleftheriou said. "The tape will be used as a deep archive."
On average, the CERN hadron collider stores 15 petabytes of data per year.
The tape cartridges IBM is developing will hold up to 128TB of data each, significantly more than today's technology, which holds about 6TB. In order to increase the capacity, technologists in IBM's Zurich Research Laboratory must increase the areal density of today's tape cartridges by a factor of 10, which Eleftheriou said his team has already proved it can do.
Eleftheriou's team has already demonstrated it can store 29.6Gb of data per square inch of magnetic tape. The team is shooting for 100Gb per square inch in order to achieve the 128TB tape cartridges it is seeking. While latency would be a problem if tapes were stored offsite, tapes stored in a robotic library will be able to be read sequentially, once the beginning of a file is reached. Today's enterprise-class tape drives afford up to 250MBps throughput, more than enough for data hungry scientists.
Construction is expected to start in 2016 and take four years. The exascale supercomputer is expected to be completed by 2024, enabling SKA to be online by 2026, Eleftheriou said.
The actual radio telescope array will be far larger than 1 kilometer. The kilometer in the SKA's name refers to the total collecting area of the SKA, which will be 1 million square meters, according to the SKA website. To achieve this, the SKA will use 3,000 dish antennas, each about 15 meters in diameter as well as two other types of radio wave receptor, know as aperture array antennas. The antennas will be arranged in five spiral arms and the dishes will extend to distances of at least 3,000 kilometers from the center of the long baseline array.
Closeups of the SKA's radio telescope dishes (Source: ASTRON)
Lucas Mearian covers storage, disaster recovery and business continuity, financial services infrastructure and health care IT for Computerworld. Follow Lucas on Twitter at @lucasmearian or subscribe to Lucas's RSS feed. His e-mail address is firstname.lastname@example.org.
Read more about data storage in Computerworld's Data Storage Topic Center.