Multi-Master database replication
Last updated
Was this helpful?
Last updated
Was this helpful?
iCure Database is built on Apache CouchDB a web enabled database optimized for use over a network with multiple users, oriented at massively distributed systems with heavy usage that handle "Big Data".
CouchDB does not store data in a table along with relationships, but uses a B-Tree instead. Each database in every user device is a collection of independent documents that maintain their own data and self-contained schemas. Document metadata contains revision information, so that differences happening while databases were disconnected, may be merged. CouchDB impelements its own version of a Multiversion Concurency Controll system, that supports .
This enables the iCure to implement a fully . Files are not locked during update and they can be updated both online and offline, allowing the iCure to implement a conflict resolution method.
Under normal conditions database synchronization between connected databases take place in less that 300ms.
CAP Theorem dictates that Databases can be optimized for two of the three most requested qualities: Consistency, Availability and Partition Tolerance. While CouchDB has been optimized for Partition Tolerance and Availability, it has the ability to provide local consistency at local databases and through incremental replication eventual consistency in the iCure database.
Other features include document-level semantics with , (incremental) MapReduce, and (incremental) replication.
The has been designed so that it is agnostic to higher level Medical Data conceptual Data Models such as Open EHR and FHIR. It does however make some basic assumptions over the nature of the data, that are to be utilized for. That is in line with the selection of CouchDb. In a healthcare setup one might expect that there are several encryption perimeters and distributed systems, that are to some extend experiencing aberrant disconnections.
