HomeArticlesSlideshowsAsk BobLogin or Register
NewsCustom SearchContactLinks
Data interfaces - independence of particular systems within enterprise
There are two basic ways to keep Information systems manageable:
  • Perfect enterprise metadata system (I dream about it but I have never seen it:)
  • Independent systems with defined services and data interfaces more realistic method
Most non success stories producing unmanageable bulks of systems come from attempts to reach the first one. So we will now talk about the second one better.

Data Interface contract - specifies following aspects:
  • Form of interface - naming, fields, data types, constraints.
  • SLA - time when data will be in place, periodicity.
  • Archiving - retention period of data in the interface.
  • Signals - optionally it is possible to define way how systems share information about state of interface (eg. READY, PROCESSED).
  • Corrections - way to put corrected information.

Issues mitigated by independence layers:
  • Process independence - low influence to source performance.
  • Data read consistency - old snapshot issue.
  • Impact analysis before changes - interface has the same manageable role as any other function of the system.
  • Changes of data structures within source system.
  • Changes of data structures within target system.
  • Changes of platform of source system.

Data interface based on Plain file extracts:
  • Maximum level of independence (including source system platform).
  • Usable as only standard way to process batch data interfaces.
  • Higher independence of load - source could be down during load.
  • Additional costs to extract/import data.
  • Additional costs to safely transport data between systems.

Data interface based on DB-link:
  • Maximum efficiency for homogeneous systems.
  • Safely works with native data types and encodings.
  • Independence of source structure solved as an extract layer (database views or output stage tables).
  • Independence of source platform partially solved using as staging area on target system - filling of staging tables will be replaced from the database link to another way - it does not touch logic of transformation process.
  • Reduces costs to transfer files between nodes.
  • Because of process independence of target and source the transport should be 1:1 to reduce time and resource consumption.

Ludek Bob Jankovsky
All Right Reserved © 2007, Designed by Bob Jankovsky