PERGAMUM: REPLACING TAPE WITH ENERGY EFFICIENT, RELIABLE, DISK-BASED ARCHIVAL STORAGE

1
PERGAMUM REPLACING TAPE WITH ENERGY EFFICIENT,
RELIABLE,DISK-BASED ARCHIVAL STORAGE

• M. W. Storer K. M. Greenan E. L. MillerUCSC
• K. VorugantNetwork Appliance

2
Summary

• These slides only include what is missing in the
  Shockwave presentation
• Background on MAID
• Glossary
• Two-level redundancy
• Digital signatures
• They are not self-sufficient

3
MAID

• Massive Array of Idle Disks
• Designed for Write Once, Read Occasionally
  applications.
• Each disk is only spun up when a request is
  issued for the data on disk
• Problems
• Less redundancy than RAID arrays
• Frequent spin ups wear disks

4
Glossary (I)

• Tome
• Basic hardware unit
• Consists of
• A low-power CPU with DRAM
• A commodity SATA drive
• A small flash drive
• An Ethernet connection
• Very autonomous

5
Glossary (II)

• Blocks
• Much larger than conventional disk blocks
• 128 KB to 1 MB
• Will batch up small files into large files for
  archival storage
• Normally done for tape drives (tar)

6
Glossary (III)

• Segments
• Consist of a fixed number of blocks
• Intra-disk parity is computed inside each segment
• Includes one or two parity blocks

7
Glossary (II)

• Regions
• Formed by a given number of segments
• Basic building block for inter-disk redundancy

8
Two-level redundancy
9
Intra-disk redundancy

• Used to fix unrecoverable read errors (bad
  blocks) without spinning up disks of other tomes
• Save energy
• Reduce frequency of disk scrubbing operations
• Can reconstitute contents of bad blocks

10
Inter-disk redundancy

• Used to recover from disk failures
• Recovery process is designed to minimize peak
  power consumption rather than total power
  consumption

11
Digital signatures

• Include a timestamp (32 bits)
• Stored in flash memory
• Used to detect
• Incorrect propagation of updates
• Write errors