As data continues to hold immense value, Apache Lucene has emerged as a leading full-text search engine renowned for its ability to search vast amounts of textual data using inverted indexing efficiently. However, being a standalone solution, Lucene poses scalability challenges as data volumes increase, requiring frequent index rebuilding that can significantly hamper performance. While Java and REST-based alternatives now offer scalable options for full-text search, the .NET stack still lacks a seamlessly integrated, scalable solution for this critical function.

Using Distributed Lucene with NCache for .NET

NCache, a powerful and popular .NET in-memory data store, has implemented native Lucene.NET API over its distributed architecture. As it is the standard Lucene.NET API, no code change is required to use it in a scalable manner with NCache.

NCache also utilizes Lucene.NET to create indexes in a dynamically scalable environment to allow distributed full-text searches. The results of these searches merge before being sent back to your application. This enhances the stand-alone Lucene into a fast, linearly scalable full-text searching solution.

This enhances the stand-alone Lucene into a fast, linearly scalable full-text searching solution.

Using Lucene in .NET Apps

Let us consider an e-commerce site that holds information on thousands of products, orders, and customer details. Hence, indexing all attributes especially non-textual fields (which are not used while searching), is not a wise approach, as it exhausts the cache memory.

For example, our document for a product looks like this:

{

“ID”: “ABC34”,

“Name”: “Tupperware”,

“Description”: “Microwaveable, dishwasher-friendly, reusable

Tupperware in three sizes”,

“RetailPrice”: 15.00,

“Discount”: 3.00

}

Now, we know that our clients perform full-text searches specifically on the product description field within documents. To optimize this process, we propose indexing only searchable fields and associating them with keys that link back to their corresponding documents in our persistence store, such as a database or file system. For instance, when querying for products like “dishwasher-friendly Tupperware,” all relevant products matching these terms will be retrieved with their ProductID serving as the document key. Subsequently, the complete document can be efficiently fetched from the persisted index, streamlining search operations.

To use Distributed Lucene in your existing applications, all you need is to specify NCacheDirectory when opening a directory. This requires the NCache cache name and the index name. The following code snippet opens a directory on a cache LuceneCache in NCache and an index named ProductIndex.

// Specify the cache name and index path for Lucene

string cache = "LuceneCache";

string index = "ProductIndex";

// Create directory and open it on the cache

Directory directory = NCacheDirectory.Open(cache, index);

Lucene ships an extensive query language, which interprets a given string into a Lucene query. This can be done either on a term, multiple terms, wildcards, or even fuzzy words. To learn more about Lucene queries, read Lucene Query Docs.

The following code snippet creates an IndexReader on the directory, which is used by the IndexSearcher. The data is analyzed and tokenized based on the StandardAnalyzer. The first 50 hits from the result are returned to the application. Note that the analyzer must be the same as the one used during index creation.

// The 'applyAllDeletes' is true so all enqueued deletes are applied on writer

IndexReader reader = DirectoryReader.Open(indexWriter, true);

// A searcher is opened to perform searching

IndexSearcher indexSearcher = new IndexSearcher(reader);

// Specify the searchTerm, fieldName and analyzer

string searchTerm = "Beverages";

string fieldName = "Category";

// Note that the analyzer should be same as the one used during index creation

Analyzer analyzer = new StandardAnalyzer(LuceneVersion.LUCENE_48);

// Create a query parser to parse the query

QueryParser parser = new QueryParser(LuceneVersion.LUCENE_48, fieldName, analyzer);

Query query = parser.Parse(searchTerm);

// Returns the top 50 hits from the result set

ScoreDoc[] docsFound = indexSearcher.Search(query, 50).ScoreDocs;

reader.Dispose();

Load Data to Build Distributed Index

With Lucene, you can build indexes and load data into them as needed. Indexes require an analyzer, that analyzes and tokenizes the data according to your need. This includes whitespace, non-letters, punctuation, and so on. Once you create a writer for your Lucene index, you can create documents and add fields to it. This document is then indexed in NCache as a distributed index once you call Commit(). For more details on Lucene analyzers, have a look at Lucene Analyzer Docs.

// Specify the cache name and index path for Lucene

string cache = "LuceneCache";

string indexPath = "ProductIndex";

// Create directory and open it on the cache

Directory directory = NCacheDirectory.Open(cache, indexPath);

// The same analyzer is used as for the reader

Analyzer analyzer = new StandardAnalyzer(LuceneVersion.LUCENE_48);

IndexWriterConfig config = new IndexWriterConfig(LuceneVersion.LUCENE_48, analyzer);

// Create indexWriter on NCache directory

IndexWriter indexWriter = new IndexWriter(directory, config);

Product[] products = FetchProductsToIndex();

foreach (var product in products)

{

Document doc = new Document

{

new StoredField("id", product.ID),

new TextField("name", product.Name, Field.Store.YES),

new TextField("description", product.Description, Field.Store.YES),

new StringField("category", product.Category, Field.Store.No),

new StoredField("retail_price", product.RetailPrice),

};

indexWriter.AddDocument(doc);

}

indexWriter.Commit();

Why NCache for Distributed Lucene?

Using NCache for Distributed Lucene provides you with the following benefits:

Extremely Fast and Linearly Scalable: NCache is an in-memory distributed data store, so building distributed Lucene on top of it provides the same optimum performance for your full-text searches. Moreover, because of NCache’s distributed architecture, the Lucene index is partitioned across all the servers of the cluster. This makes it scalable as you can add more servers on the go as your data load increases, and Lucene indexes are automatically redistributed without any client intervention.

Data Replication for Reliability and High Availability: With NCache’s Partition-Replica topology, the Lucene index is not only partitioned across all servers but each partition is also replicated to another server of the cluster. Hence, if any server goes down, the replica of the partition serves all the queries for that index, ensuring reliability. Similarly, if a server node goes down, NCache dynamically self-heals by readjusting the data within the remaining nodes, without any downtime or impact to your Lucene index, ensuring high availability.

Conclusion

To sum it up, full-text searching has now become fundamental in almost every business, owing to the powerful search engine Lucene. But as data grows, rebuilding indexes can cause more damage than gain and this is where an in-memory, distributed .NET solution such as NCache steps in. All it requires is a one-line code change in your existing Lucene application, and voila, you have the best of both worlds as an in-memory, distributed full-text search mechanism.