Nearly two and a half years after announcing it would digitize its entire film and radio archive, the Australian Broadcasting Corporation's data conversion work starts this month.

The ABC's systems management services manager Fred Spark told Computerworld that due to the scale of the project "it took some time to convince the market that they could deliver our vision".

During the past nine months the systems on which the conversion of the archive will operate have been installed, including a national rollout of IBM servers, storage arrays, and tape libraries.

"As it's a one-for-one, read-and-write process it is very labour intensive," Spark said, adding that a workshop has been set up for three shifts of seven people to do the conversion.

The scale of the project is likely to set an Australian precedent for digital storage as Spark estimates the digital archive will grow, without allowing for new content, to 1.5 petabytes in three to five years.

The original tender for the system was $10 million but by the time the conversion begins, Spark said additional services requirements will blow that out to around $15 million.

"The archive is all tape and is made up of analog and digibeta," Spark said. "All of it will be ingested and translated to LTO." Spark said tape is the best option for an archive of this size, and the data will be readily accessible for producers to retrieve footage and reports "in minutes rather than hours or days".

"It's like disk response times [and] LTO won't age as quickly as disk," he said.

The infrastructure consists of a pSeries machine running the Ardendo archiving software in each capital city, FastT storage arrays in Sydney and Melbourne, and tape libraries also in every capital. So far 2TB of disk and 500TB of tape has been installed.

"The pSeries machines range from 520s in Sydney and Melbourne down to smaller machines in the other states," Spark said. "High-end pSeries machines are expensive but the smaller boxes very competitive but still expensive compared with Linux on Intel."

Spark is confident the performance of the Power-based pSeries machines is up to the task saying, "these things fly".

For its corporate information, the ABC has about 18TB of disk (half SATA) in its Hitachi Data Systems SAN and is looking at virtualization to better utilize its capacity, in addition to replicating half of that capacity to a HDS-based DR site.

"Virtualization is certainly on the ABC's agenda," Spark said. "Storage is moving towards virtualization and so far our strategy has been untidy because of [the] number of initiatives."