You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 15 Next »

Requirements


  1. Allow user to create Lucene Indexes on data stored in Geode

  2. Update the indexes asynchronously to avoid impacting write latency
  3. Allow user to perform text (Lucene) search on Geode data using the Lucene index. Results from the text searches may be stale due to asynchronous index updates.

  4. Provide highly available of indexes using Geode's HA capabilities 

  5. Scalability
  6. Performance comparable to RAMFSDirectory

Out of Scope
  1. Building next/better Solr/Elasticsearch.

  2. Enhancing the current Geode OQL to use Lucene index.

Related Documents

A previous integration of Lucene and GemFire:

   

Similar efforts done by other data products

  • Hibernate Search: Hibernate search

  • Solandra: Solandra embeds Solr in Cassandra.

Terminology

  • Documents: In Lucene, a Document is the unit of search and index. An index consists of one or more Documents.
  • Fields: A Document consists of one or more Fields. A Field is simply a name-value pair.
  • Indexing involves adding Documents to an IndexWriter, and searching involves retrieving Documents from an index via an IndexSearcher.

API

User Input

  1. A region and list of to-be-indexed fields
  2. [ Optional ] Specified Analyzer for fields or Standard Analyzer if not specified with fields

Key points

  1. A single index will not support multiple regions. Join queries between regions are not supported

  2. Heterogeneous objects in single region will be supported
  3. Only top level fields and nested objects can be indexed, not nested collections
  4. The index needs to be created before adding the data (for phase1) 
  5. Pagination of results will be supported

Users will interact with a new LuceneService interface, which provides methods for creating indexes and querying. Users can also create indexes through gfsh or cache.xml.

Java API 

LuceneService

/**
   * Create a lucene index using default analyzer.
   */
  public LuceneIndex createIndex(String indexName, String regionName, String... fields);
  
  /**
   * Create a lucene index using specified analyzer per field
   */
  public LuceneIndex createIndex(String indexName, String regionName,  
      Map<String, Analyzer> analyzerPerField);

  public void destroyIndex(LuceneIndex index);
 
  public LuceneIndex getIndex(String indexName, String regionName);
  
  public Collection<LuceneIndex> getAllIndexes();

  /**
   * Get a factory for building queries
   */ 
  public LuceneQueryFactory createLuceneQueryFactory();
  

LuceneQueryFactory

public enum ResultType {
    /**
     *  Query results only contain value, which is the default setting.
     *  If field projection is specified, use projected fields' values instead of whole domain object
     */
    VALUE,
    
    /**
     * Query results contain score
     */
    SCORE,
    
    /**
     * Query results contain key
     */
    KEY
  };
 /**
   * Set page size for a query result. The default page size is 0 which means no pagination.
   * If specified negative value, throw IllegalArgumentException
   * @param pageSize
   * @return itself
   */
  LuceneQueryFactory setPageSize(int pageSize);
  
  /**
   * Set max limit of result for a query
   * If specified limit is less or equal to zero, throw IllegalArgumentException
   * @param limit
   * @return itself
   */
  LuceneQueryFactory setResultLimit(int limit);
  
  /**
   * set weather to include SCORE, KEY in result
   * 
   * @param resultTypes
   * @return itself
   */
  LuceneQueryFactory setResultTypes(ResultType... resultTypes);
  
  /**
   * Set a list of fields for result projection.
   * 
   * @param fieldNames
   * @return itself
   */
  LuceneQueryFactory setProjectionFields(String... fieldNames);
  
  /**
   * Create wrapper object for lucene's QueryParser object.
   * The queryString is using lucene QueryParser's syntax. QueryParser is for easy-to-use 
   * with human understandable syntax. 
   *  
   * @param regionName region name
   * @param indexName index name
   * @param queryString query string in lucene QueryParser's syntax
   * @param analyzer lucene Analyzer to parse the queryString
   * @return LuceneQuery object
   * @throws ParseException
   */
  public LuceneQuery create(String indexName, String regionName, String queryString, 
      Analyzer analyzer) throws ParseException;
  
  /**
   * Create wrapper object for lucene's QueryParser object using default standard analyzer.
   * The queryString is using lucene QueryParser's syntax. QueryParser is for easy-to-use 
   * with human understandable syntax. 
   *  
   * @param regionName region name
   * @param indexName index name
   * @param queryString query string in lucene QueryParser's syntax
   * @return LuceneQuery object
   * @throws ParseException
   */
  public LuceneQuery create(String indexName, String regionName, String queryString) 
      throws ParseException;
  
  /**
   * Create wrapper object for lucene's Query object.
   * Advanced lucene users can customized their own Query object and directly use in this API.  
   * 
   * @param regionName region name
   * @param indexName index name
   * @param query lucene Query object
   * @return LuceneQuery object
   */
  public LuceneQuery create(String indexName, String regionName, Query query);

LuceneQuery

/**
 * Provides wrapper object of Lucene's Query object and execute the search. 
 * <p>Instances of this interface are created using
 * {@link LuceneQueryFactory#create}.
 * 
 */
public interface LuceneQuery {
  /**
   * Execute the search and get results. 
   */
  public LuceneQueryResults<?> search();
  
  /**
   * Get page size setting of current query. 
   */
  public int getPageSize();
  
  /**
   * Get limit size setting of current query. 
   */
  public int getLimit();
  /**
   * Get result types setting of current query. 
   */
  public ResultType[] getResultTypes();
  
  /**
   * Get projected fields setting of current query. 
   */
  public String[] getProjectedFieldNames();
}
 

LuceneResultStruct

  /**
   * Return the value associated with the given field name
   *
   * @param fieldName the String name of the field
   * @return the value associated with the specified field
   * @throws IllegalArgumentException If this struct does not have a field named fieldName
   */
  public Object getProjectedField(String fieldName);
  
  /**
   * Return key of the entry
   *
   * @return key
   * @throws IllegalArgumentException If this struct does not contain key
   */
  public Object getKey();
  
  /**
   * Return value of the entry
   *
   * @return value the whole domain object
   * @throws IllegalArgumentException If this struct does not contain value
   */
  public Object getValue();
  
  /**
   * Return score of the query 
   *
   * @return score
   * @throws IllegalArgumentException If this struct does not contain score
   */
  public Double getScore();
  
  /**
   * Get the types of values ordered list
   * Item in the list could be either ResultType, or field name
   * @return the array of result types
   */
  public Object[] getNames();
  
  /**
   * Get the values in same order as result types
   * @return the array of values
   */
  public Object[] getResultValues();
}

    Examples

// Get LuceneService
LuceneService luceneService = LuceneServiceProvider.get(cache);

// Create Index on fields with default analyzer:
LuceneIndex index = luceneService.createIndex(indexName, regionName, "field1", "field2", "field3");

// create index on fields with specified analyzer:
LuceneIndex index = luceneService.createIndex(indexName, regionName, analyzerPerField);

// Create Query
LuceneQuery query = luceneService.createLuceneQueryFactory().setLimit(200).setPageSize(20)
  .setResultType(SCORE, VALUE, KEY).setFieldProjection("field1", "field2")
  .create(indexName, regionName, querystring, analyzer);

// Search using Query
LuceneQueryResults results = query.search();

List values = results.getNextPage(); // return all results in one page

// Pagination
while (results.hasNextPage())
  List page = results.getNextPage(); // return result page by page

  for (LuceneResultStruct r : page) {
    System.out.prinlnt(r.getValue());
  }
}

 

Gfsh API

 

// Create Index
gfsh> create lucene-index --name=indexName --region=/orders --fields=customer,tags

// Destory Index
gfsh> destroy lucene-index --name=indexName --region=/orders

Execute Lucene query
gfsh> luceneQuery --regionName=/orders -queryStrings="" --limit=100 page-size=10

 

XML Configuration 

 

<region name="region">  
 <lucene-index indexName="luceneIndex">
             <FieldDefinition name="fieldName" analyzer="KeywordAnalyzer"/> 
 </lucene-index>
</region>

 

REST API

TBD - But using solr to provide a REST API might make a lot of sense

Spring Data GemFire Support

TBD - But the Searchable annotation described in this blog might be a good place to start.

Implementation Flowchart

 

 

Index Storage

The lucene indexes will be stored in memory instead of disk. This will be done by implementing a lucene Directory called RegionDirectory which uses Geode as a flat file system. This way we get all the benefits offered by Geode and we can achieve replication and shard-ing of the indexes. The lucene indexes will be co-located with the data region in case of HA. 
A LuceneIndex object will be created for each index, to manage all the attributes related with the index, such as reflection fields, AEQ listener, RegionDirectory array, Search, etc. 

  Colocated PR or Replicated Region User Data Region Async Queue Lucene Regions LuceneIndex RegionDirectory User Puts Batch Writes

Inside LuceneIndex

LuceneIndex Reflective fields AEQ listener RegionDirectory array (one per bucket) Query objects 

A closer look at Partitioned region data flow

LuceneIndex AEQ listener processes events into index documents RegionDirectory1 file region bucket 1 chunk region bucket 1 RegionDirectory2 file region bucket 2 chunk region bucket 2 User User Data Region Bucket 1 Async Queue Bucket 1 Bucket 2 Async Queue Bucket 2 PUTs Batch Write Batch Write


If user's data region is a partitioned region, there will be one LuceneIndex is for the partitioned region. Every bucket in the data region will have its own RegionDirectory (implements Lucene's Directory interface), which keeps the FileSystem for index regions. Index regions contain 2 regions:
  • FileRegion : holds the meta data about indexing files
  • ChunkRegion : Holds the actual data chunks for a given index file. 

The FileRegion and ChunkRegion will be collocated with the data region which is to be indexed. The FileRegion and ChunkRegion will have partition resolver that looks at the bucket id part of the key only.
In AsyncEventListener, when a data entry is processed
  1. create document for indexed fields
  2. determine the bucket id of the entry.
  3. Get the RegionDirectory for that bucket, save the document into RegionDirectory. 

Storage with different region types

PersistentRegions
The Lucene Index will be persisted.
OverflowRegions
The Lucene Index will not be overflowed. The rational here is that the Lucene index will be much smaller than the data size, so it is not necessary to overflow the index.
EmptyRegions
The Lucene Index not supported
OffHeapRegions
The Lucene index will be stored in OffHeap

Index Maintenance

An AsynchEventQueue will be used to update the LuceneIndex. This will allow us to do updates in batches supported by AEQ. Indexed field values are obtained from AsynchEvent through reflection (in case of domain object) or by PdxInstance interface (in case pdx or JSON); constructing Lucene document object and adding it to the LuceneIndex associated with that region.

 

Handling failures, restarts, and rebalance 

The index and async event queue will be stored and a region with the same redundancy level as the original region. We will take care to ensure that all updates are written to the index files before removing events from the queue. So during failover the new primary should be able to read index files from disk.

 

Walkthrough creating index in Geode region

  
1) Create a LuceneIndex object to hold the data structures that will be created in following steps. This object will be registered to cache owned LuceneService later. 
2) LuceneIndex will keep all the reflective fields. 
3 Assume the dataregion is PartitionedRegion (otherwise, no need to define PartitionResolver). Create a FileRegion (let's call it "fr") and a ChunkRegion (let's call it "cr"), collocated with Data Region (let's name it "dataregion"). FileRegion and ChunkRegion use the same region attributes as dataregion. In partitioned region case, the FileRegion and ChunkRegion will be under the same parent region, i.e. /root in this example. In replicated region case, the index regions will be root regions all the time. 
 
3) Create a GeodeDirectory object using the FileRegion, ChunkRegion and the path we got in previous step. 
 
4) Create PerFieldAnalyzerWrapper and save the fields in LuceneIndex. 
 
5) Create a Lucene's IndexWriterConfig object using Analyzer. 
 
6) Create a Lucene's IndexWriter object using GeodeDirectory and IndexWriterConfig object. 
 
7) Define PartitionResolver to use dataregion's bucket id as routing object, which will guarantee the index bucket region will be the same bucket id as the dataregion's bucket region's even when dataregion has its own customer-defined PartitionResolver. We don't nedd to define PartitionResolver on dataregion. 
 
8) Define AEQ with multiple dispatcher threads and order-policy=partition. That will group events by bucket id into different dispatcher queues. Each dispatcher thread will call our AEQ listener to process events for one or more buckets. Each event will be processed to be Document and write into ChunkRegion via GeodeDirectory. We don't need lock for GeodeDirectory, since only one thread will process one bucket's events. 
 
9) If dataregion is a replicated region, then define AEQ with single dispatcher thread. 
 
10) Register the newly created LuceneIndex into LuceneService. The registration step will also publish the meta data into the "lucene_meta_region" which is a persistent replicate region, then other JVM will know a new luceneIndex with these meta data was created. All the members should have a LuceneService instance with the same LuceneIndex definition.

Processing Queries
 

Partitioned regions

In the case of partitioned regions, the query must be sent out to all of the primaries. The results will then need to be aggregated back together. We are still investigating options for how to aggregate the data, see Text Search Aggregation Options.

Replicated regions

TBD


Result collection and paging

The ResultSet will support pagination mechanism to retrieve the results. All the keys are aggregated at the query executor node (client or peer); and getAll is used to fetch the values according to page size.
  • No labels