Archival Storage
The Petabyte Storage Infrastructure Project includes multiple
components, for data collection (motes), disk space for data processing
(fast disk cache), and long term data storage (archival storage). This
section is devoted to the latter aspect of the project.
Year 1
As we are approaching the end of year one, we are in the process of making our
initial bulk purchase of Storage Bricks.
The final model consists of 1TB of PATA disks in a low power, 1U form factor.
The initial Archival Storage Network
infrastructure is now in place. We have targeted three separate machine rooms
allowing for a three-way mirroring of the data. All of the current equipment
is 1000bT capable so that we can evaluate the differences between 100bT
and 1000bT to the individual storage nodes.
The Archival Software is less well
developed at this point, although much brainstorming and some
experimentation has been done to probe the design space. We have
examined various redundancy and hardware replacement strategies. We plan on
using erasure
codes for redundancy and hot-spares plus swapping boxes
after a second disk failure for hardware replacement.
See Also
|