In the search for the right technology, I identified a compromize between SQL based datastorage and flat files.
even so, parameters are many and multidimensional, also genetically evolving - they are not massive, like test results so the performance of a SQL db is not required.
in opposition, what is required and seems to be covered by NoSQL database approaches like CouchDB. http://wiki.apache.org/couchdb/
- autrak server to server communication. not always - on philosophy allows redundancy and loose connection between systems but being able to use a common repository.
- easy changing structure of storage. strategies with an evolutionary approach develop own models and therefore extend or reduce dimensions of data. this means - the storage model is dynamic. that is more important than high-throughput.
- Mutli version concurrency control is the power I need here.
- the google patented Map-reduce method helps to maintain still performance.
- consistency and conflicts between storages can be covered by the system.
- REST protocoll based JSON-API which is close to other service based storages.
- document oriented file management system.
- accessible from all available programming languages and OS.
- functional programming language Erlang is used.
- master/slave replication and caching of remotely managed data.
- docuemnts have a unique ID
- queries can traverse any document which has the correct fieldname inside. anything else does not matter. - very flexible for changing structure.
- new views can be created from query results, and simply create new documents with the combined results. this makes it easy to summarize and merge different data and structures without excessive relational database modelling and mapping.
- distributed database system and fully redundant
- open source and cloud based.
final summary - fucking simple to use and really powerful. core technolgy used by Google, Amazon & Co
even so, parameters are many and multidimensional, also genetically evolving - they are not massive, like test results so the performance of a SQL db is not required.
in opposition, what is required and seems to be covered by NoSQL database approaches like CouchDB. http://wiki.apache.org/couchdb/
- autrak server to server communication. not always - on philosophy allows redundancy and loose connection between systems but being able to use a common repository.
- easy changing structure of storage. strategies with an evolutionary approach develop own models and therefore extend or reduce dimensions of data. this means - the storage model is dynamic. that is more important than high-throughput.
- Mutli version concurrency control is the power I need here.
- the google patented Map-reduce method helps to maintain still performance.
- consistency and conflicts between storages can be covered by the system.
- REST protocoll based JSON-API which is close to other service based storages.
- document oriented file management system.
- accessible from all available programming languages and OS.
- functional programming language Erlang is used.
- master/slave replication and caching of remotely managed data.
- docuemnts have a unique ID
- queries can traverse any document which has the correct fieldname inside. anything else does not matter. - very flexible for changing structure.
- new views can be created from query results, and simply create new documents with the combined results. this makes it easy to summarize and merge different data and structures without excessive relational database modelling and mapping.
- distributed database system and fully redundant
- open source and cloud based.
final summary - fucking simple to use and really powerful. core technolgy used by Google, Amazon & Co
No comments:
Post a Comment