Flashcards
Data Management
Cards in this deck (46)
- 1Conceptual schemaTechnology independent specification of the overall structure of organisational data
- 2Logical schemaRepresentation of a database for a particular data management technology
- 3Physical schemaHow data from a logical schema are stored
- 4Weak entity typeAn entity whose existence depends on some other entity type
- 5Composite attributeAttribute that has meaningful parts
- 6IdentifierAttribute that distinguishes instances of an entity type
- 7Degree of relationshipNumber of entity types that participate in a relationship
- 8Cardinality constraintsA rule that specifies the number of instances of one entity
- 9Total specialization ruleEach instance of a supertype must be a member of some subtype in the relationship
- 10Partial specialization ruleInstance of a supertype may be member of a subtype but dont have to
- 11Disjoint ruleA supertype instance cannot be a member of both subtypes
- 12Overlap ruleSupertype instance may simutaneously be member of 2 subtypes
- 13Entity ClusteringHow Supertypes & Subtypes
Related entity types & relationships - 14Candidate keyattribute which uniquely identifies rows in a relation
- 15Characteristics Datawarehouse1) subject oriented: (organised around key subjects)
2) Integrated
3) time-variant: trends & situations - 16Difference operational & informational systemOperational: transactions
Informational: decision-support, predictions, data-mining - 17Independent datamartData filled with operational data without beingc onnected to DWH
- 18Limitations independent DMcostly, redundant, inconsistencies, no greater details
- 19Operational Data Store (ODS)Database designed to serve operational user in decision-support (recent data)
- 20Logical data martDM offering different relational views
- 21Real-time WHnear real-time data - immediate action can be taken in response to events
- 22Three-layer architecture DW1) Derived data (end-user decision-support)
1) Reconciled data (decision-support)
3) Operational data - 23Fact table vs. dimensional tablesFactual quantitative data (units sold, orders booked)
subject of business - 24Size of a fact tablesubjects (stores * products) * time
- 25Weight fact tableSize * fields * bytes
- 26SCDType 1: overwrite
Type 3: current value & many old - 27Type 2 methodDimensions that change slowlier (address) PK +Start date (End date- automatically)
- 28Distributed databasesingle logical database spread physically across computers in multiple locations & connected by a data communication network
*centrally administrated, but local customization & flexibility - 29Decentralized databaseDB stored on multiple locations but computers are not interconnected by network - users cannot share data
- 30Advantages distributed databases*Distribution & autonomy of business units (geographically dispersed units)
*Data sharing (Sharing across business units)
*Data communication costs & reliability (locate data close to where it is needed)
*Database recovery (ensuring that damaged data can be quickly recovered) - 31ObjectivesLocation transparency - a user needs to know the location of the data
Local autonomy - ability to operate independently when the connection to other nodes has failed - 32Synchronous Distributed DatabaseData is continuously kept up-to-date
*Data can be accessed at any time and provide consistent answers
*lead to slow response time - 33Asynchronous DD*Keeps copies of replicated data at different nodes
*Local server can access data without reaching out across the network
*Delay in updates
*The response time is acceptable
--> more complex to design to ensure the suitable level of data integrity and consistency across nodes - 34Advantages DD over centralized*Increased reliability and availability (when centralized data fails- unavailable to all users)
*Local control (thus data integrity)
*Modular growth (more economical than expand to larger computer)
*Lower communication point (where needed)
*Faster response - 35Disadvantages DD*Software costs & complexity
*Complex to ensure proper coordination among data at different sites - 36Options for distributing a database1) Data replication
2) Horizontal partitioning
3) Vertical partitioning - 37Data replication*Storing a separate copy of the DB at 2+ sites
*Every site = full replication
*Favored when data is static - 38Advantages DR*Reliable: always a copy
*Fast response (process query locally)
*Note decoupling (each node proceed without coordination across network)
*Reduced network traffic at prime time (delayed updating thus traffic moved away) - 39Disadvantages*Storage requirements
*Complexity maintenance
*Complexity & cost of updating - 40Data replication applications*Snapshot replication (Updates periodically collected, list of changes in log)
*Near-real time (Trigger stored at DB, when replicated, trigger executes update)
*Pul replication (Local DB determines when update & requests snapshot) - 41Horizontal partitioningRows distributed to many sites
- 42Advantages HP*Efficiency: data stored close to where it is used
*Local optimization (optimize performance for local access)
*Security (data not relevant to usage)
Ease of querying (combining data easy - rows merged by unions) - 43Disadvantages HP*Inconsistent access speed (when data from different partitions are needed)
*Backup vulnerability: no copies -data lost if backup is not performed - 44Vertical partitioningColumns
- 45Disadvantages VP*combining data more difficult than horizontal patterns (due to need to use PK to join rows across partitions)
- 46Objectives DBMS*Location transparency (data dictionary indicating location of data)
*Replication (although replicated, data should be treated as single item)
*Failure transparency (either all actions commited or none)
*Concurrency transparency (results must be the same as if each transaction were processed in serial order)
Information
- Description
- Data Management Flashcards on Data Management, created by L H on 19/12/2016.
- Created by
- L H
- Views
- 3
- Rating