NGDA Architecture Update

Greg Janée

Greg Janée • May 16, 2005 2

Three motivations

• Archival has to be cheap & easy– little incentive– no funding

• Need to archive data semantics– key differentiator from text, audio, video

• Focus on long-term preservation– need to migrate whole systems

Greg Janée • May 16, 2005 3

system

databasestorage

handleresolver

database

Typical repository architecture

database

handleresolver

database

fragile

Greg Janée • May 16, 2005 4

NGDA architecture

storage subsystem

standard, public data model

archival system

ADL OAIbulk

loader

databases,caches,

etc.

Web

access ingest

Greg Janée • May 16, 2005 5

Post-NGDA architecture

storage subsystem

standard, public data model

Web

Greg Janée • May 16, 2005 6

Storage system requirements

• Req’s:– associate UUIDs/RIDs with bitstreams– retrieve global/local bitstream by UUID/RID– determine (parent) UUID of any bitstream– list all UUIDs

• Satisfied by:– any filesystem– tag URIs for UUIDs

• tag:library.ucsb.edu,2005:identifier

Greg Janée • May 16, 2005 7

Archival objects

directoryUUID

componentRID

UUID

Greg Janée • May 16, 2005 8

Archival objects

• Directory info per component– named relationship/position– format & semantics

• by UUID references to definitions

– fixity: checksum– provenance: isDerivative– policy: mutability– rights

• Components may be provided by archive itself

Greg Janée • May 16, 2005 9

Example

USGS

DOQQ

GeoTIFFFGDC

Object x

x.tiffx.fgdc x.gif

met

adat

ad

ata

derived

TIFFsubtypeOf

Greg Janée • May 16, 2005 10

Archives

• Archive = set of archival objects– no structure– no free-floating bitstreams

• In anticipation of federation:– associations may cross archive boundaries– archival objects may not

Greg Janée • May 16, 2005 11

Object types

• Content• Format definition• Semantic definition• Provider• Organizational structures

– collection– series– ingest session

Greg Janée • May 16, 2005 12

Archive-provider agreement

• Defines– common structure of objects to be ingested– necessary validations– associations to other objects– policies, rights, etc.

• Represents choke point– requires human evaluation

Greg Janée • May 16, 2005 13

Deferred functionality

• Incremental ingest• Object revisions• Rights• 3rd-party access• Federation

Greg Janée • May 16, 2005 14

Status

• Starting development now

• Approach: iterative refinement