Changes between Version 55 and Version 56 of ReleaseNotes


Ignore:
Timestamp:
30.01.2013 18:31:34 (12 years ago)
Author:
jri
Comment:

4.0.14 release notes, pt. 2

Legend:

Unmodified
Added
Removed
Modified

  • ReleaseNotes

    v55 v56  
    77== 2013-1-?: DeepaMehta 4.0.14 == 
    88 
    9 Improvements: 
    10 * For better performance with large amounts of data the storage layer is completely rewritten. Traversal is speedup by a new Lucene association index (#389). 
    11 * Association user data is indexed as well (analogous to topic user data) (#389). 
    12 * More performant architecture: A storage implementation is coded directly to the DeepaMehta model API. The additional bridging layer is dropped. The concept of a DM-independant MehtaGraph API is abolished. There is no dependency on the external Neo4j MehtaGraph library anymore (#389).  
    13 * Modular storage layer: a new service provider interface (`de.deepamehta.core.storage.spi.DeepaMehtaStorage`) allows 3rd-party developers to implement the DeepaMehta storage layer based on alternate DBs (besides Neo4j). Storage implementations are deployed as OSGi services (#391). 
    14 * No transactions are created anymore for DB read operations. This may yield a performance benefit (#406). 
    15 * Delivering core events involves no runtime reflection anymore. This may yield a performance benefit (#352). 
    16 * AssociationDefinition API is simplified and redundancy is removed from serialization format (#393). 
    17 * Association's Instantiation associations can be queried by type (#338). 
     9Main focus of !DeepaMehta 4.0.14 is better **Performance** for large data amounts. Read about the [[https://lists.berlios.de/pipermail/deepamehta-users/2013-January/000404.html|problem background]] in this posting. 
     10 
     11!DeepaMehta 4.0.14 comes with a completely rewritten storage layer. 
     12 
     13Main aspects of the new storage layer: 
     14* **High-speed traversal:** Traversal is significantly speed up by utilizing a Lucene index for association metadata. In particular, highly connected nodes ("supernodes") have no negative impact on traversal speed anymore (#389). 
     15* **Compact architecture:** The concept of a DM-independant !MehtaGraph API is abolished. A storage implementation is coded directly to the !DeepaMehta model API. The additional bridging layer is dropped. There is no dependency on the external Neo4j !MehtaGraph library anymore (#389).  
     16* **Modular storage layer:** a new service provider interface (`de.deepamehta.core.storage.spi.DeepaMehtaStorage`) allows 3rd-party developers to implement an alternate !DeepaMehta storage layer, based on another DB than Neo4j. Storage layer implementations are deployed as OSGi services (#391). 
     17* Association user data is indexed as well (like topic user data) (#389). 
     18 
     19Additional performance measures: 
     20* For DB read operations no transactions are created anymore (#406). 
     21* Delivering core events involves no runtime reflection anymore (#352). 
     22 
     23Further improvements: 
     24* The `AssociationDefinition` API is simplified and redundancy is removed from the serialization format (#393). 
    1825* Neo4j is updated from 1.8 to 1.8.1 
    1926 
    20 Bug Fixes: 
    21 * This exception should no longer be thrown (#392): 
    22 `java.lang.IllegalArgumentException: Reference error: ID 1234 refers to a MehtaEdge when the caller expects a MehtaNode` 
     27Bug fixes: 
     28* This exception should no longer occur (#392): 
     29  `java.lang.IllegalArgumentException: Reference error: ID 1234 refers to a MehtaEdge when the caller expects a MehtaNode` 
     30* Association's `Instantiation` associations can be queried by type (#338). 
     31 
     32Updating from previous !DeepaMehta versions is not supported. 
     33 
     34See the [[https://github.com/jri/deepamehta#readme|README]] for installation instructions. 
    2335 
    2436----