Topic: temporal database

topics > computer science > Group: database

CADES structural modeling with holons
database transactions
file system reliability
immutable files and data
persistent data structure
read-only and write-once file systems
revision delta
temporal relationships
version control


A temporal database stores data as it changes over time. A temporal or time slice query retrieves the data as of a particular time.

Variations include a rollback database and historical database. A file system may allow rollback to an earlier time.

For implementation, consider a time-split B-tree. Order-dependent queries include temporal data, while being useful for trend analysis and non-temporal applications. (cbb 8/06)

Subtopic: temporal database up

Quote: a temporal database can view tuples as seen from some time; captures history of retroactive/proactive change; sequence of historical states [»snodR9_1986]
Quote: a rollback database contains immutable snapshots indexed by time; can rollback the database to some time [»snodR9_1986]
Quote: a historical database records a single historical state per relation; supports valid time, i.e. current knowledge about the past [»snodR9_1986]
Quote: a history object is a directed graph of configuration objects; shows the history of a conceptual object [»walpJ6_1988]
Quote: a transaction object records a transaction in process; when committed, it becomes a configuration object in the naming structure (history object) [»walpJ6_1988]

Subtopic: time-slice query up

Quote: a transaction-time database answers time-slice queries concerning the state of the database at a particular time [»lomeD_1993]
Quote: time domain addressing for transactions in immutable systems; two timestamps per version for created and last read [»walpJ6_1988]

Subtopic: time-split B-tree up

Quote: with time-split b-trees, read-only transactions only need a timestamp for concurrency control; ignores all later records [»lomeD6_1989]
Quote: use time-split B-tree (TSB-tree) for time-slice queries of a transaction-time database; data clustered by time-slice and key; replicate historical versions as needed; store on WORM disk [»lomeD_1993]
Quote: time-split B-trees are space and time efficient; the fraction of redundant records is bounded [»lomeD_1993]

Subtopic: order-dependent up

Quote: an order-dependent query depends on the order of the input records; e.g., a stock's five-price moving average [»lernA9_2003]
Quote: AQuery supports ordered data as a natural property; order-of-magnitude improvement
Quote: AQuery has order-preserving and order-cavalier operators over arrables (array-table) [»lernA9_2003]
Quote: AQuery matches ASSUMING ORDER in a query with ORDERED BY in an arrable table
Quote: AQuery variables and operators refer to arrays (a column); e.g., price - prev(price) is daily-change-in-price [»lernA9_2003]

Subtopic: time series analysis up

Quote: time series analysis should be linear in time and sublinear in space; use data reduction/transformation, indexing methods, and flexible similarity searching [»shasD_2004]
Quote: StatStream monitors 30,000 time series data streams for statistics and highly correlated pairs; use DFT and grid files; easily parallelized [»shasD_2004]

Subtopic: immutable data up

Quote: consistency in immutable systems is a (named) consistent group of versions of related conceptual objects [»walpJ6_1988]
Quote: historical versions of a transaction-time database are never modified; store on WORM optical disk
Quote: Cosmos has immutable objects that are transformed into new objects by atomic actions [»walpJ6_1988]

Subtopic: backup up

Quote: use a transaction-time database for backup by time-splitting modified nodes [»lomeD_1993]

Subtopic: timestamps up

Quote: implement transaction-time database with commit-time timestamps and two-phase commit to agree on the transaction time; no extra message overhead [»lomeD_1993]
Quote: instead of using end-of-file, introduce end-of-day markers to initiate once a day processing [»jackMA5_1976]

Subtopic: database not historical up

Quote: databases may model reality but they ignore time-varying aspects, e.g. historical queries and trend analysis [»snodR9_1986]
Quote: conventional databases can not handle historical queries, trend analysis, retroactive changes, or proactive changes [»snodR9_1986]

Related Topics up

Topic: archives (19 items)
Topic: CADES structural modeling with holons (24 items)
Topic: database transactions (27 items)
Topic: file system reliability (26 items)
Topic: immutable files and data (59 items)
Topic: persistent data structure (37 items)
Topic: read-only and write-once file systems (8 items)
Topic: revision delta (18 items)
Topic: temporal relationships (40 items)
Topic: timestamps (19 items)
Topic: version control
(34 items)

Updated barberCB 2/05
Copyright © 2002-2008 by C. Bradford Barber. All rights reserved.
Thesa is a trademark of C. Bradford Barber.