Distributed system for reading RSS/Atom feeds. The system reads feeds, parses them and saves new entries into database. It also pulls the full content of the entry into database. The system is horizontally scalable (workers and multiple threads per worker) and resiliant to partial outages (using message broker).

This project was done for a challenge which was organised by Zemanta and Faculty of Computer and Information Science, University of Ljubljana. More about this challenge on official Zemanta page and on faculty page (only in Slovenian language). The project was selected as the best solution in the category of distributed web page aggregation and has consequently won the challenge in that category (news in English, news in Slovenian).

This solution requires the following systems:

• Java SE Runtime Environment 7 The logic of the system is written in JAVA programming language.
• MongoDB The MongoDB database for storing feeds and entries. Version 2.4.9
• Apache ActiveMQ Message broker for distributing workload. Version 5.9.1

Project uses Maven to define dependencies on third party libraries. Nonetheless here is the list of required libraries:

• ROME Library for RSS/Atom in Java.
• Mongo Java Driver Java driver for MongoDB.
• ActiveMQ Java driver for ActiveMQ message broker.
• JUnit Library for Unit testing.
• Joda-Time Library for Java DateTime.
• HttpComponents Library for efficent HTTP support in Java.
• Log4J Logging framework.
• Commons Codec Support for efficent encoders (e.g. SHA-1 digest util).
• Commons CLI API for parsing command line options.

The general solution consists of three JAR files:

• InsertResources: For inserting RSS feeds from CSV file into MongoDB. The CSV file consists of URLs of feeds.
• RSSDelegateWorker: For inserting jobs (feeds) into message queue and checking for stalled jobs.
• RSSMainWorker: For running thread workers which fetch entries of feeds, fetch the web page and persist it to the MongoDB. The main worker deques the job from message queue and allocates a new thread from thread pool for each feed. The thread worker then does the rest of the job.

Even though MongoDB is a schemaless database we can get a sense of application's schema, as well as any outliers to that schema using Variety, a Schema Analyzer for MongoDB.

1. Collection feeds:

   { "_id" : { "key" : "_id" }, "value" : { "type" : "ObjectId" }, "totalOccurrences" : 10000, "percentContaining" : 100 }
   { "_id" : { "key" : "accessedAt" }, "value" : { "type" : "Date" }, "totalOccurrences" : 10000, "percentContaining" : 100 }
   { "_id" : { "key" : "feedUrl" }, "value" : { "type" : "String" }, "totalOccurrences" : 10000, "percentContaining" : 100 }
   { "_id" : { "key" : "title" }, "value" : { "type" : "String" }, "totalOccurrences" : 9308, "percentContaining" : 93.08 }
   { "_id" : { "key" : "entries" }, "value" : { "type" : "Array" }, "totalOccurrences" : 9293, "percentContaining" : 92.93 }
   { "_id" : { "key" : "link" }, "value" : { "type" : "String" }, "totalOccurrences" : 9282, "percentContaining" : 92.82000000000001}
   { "_id" : { "key" : "description" }, "value" : { "type" : "String" }, "totalOccurrences" : 9189, "percentContaining" : 91.89 }
   { "_id" : { "key" : "pubDate" }, "value" : { "type" : "Date" }, "totalOccurrences" : 8205, "percentContaining" : 82.05 }
   { "_id" : { "key" : "language" }, "value" : { "type" : "String" }, "totalOccurrences" : 8003, "percentContaining" : 80.03 }
   { "_id" : { "key" : "image" }, "value" : { "type" : "Object" }, "totalOccurrences" : 4129, "percentContaining" : 41.29 }
   { "_id" : { "key" : "image.url" }, "value" : { "type" : "String" }, "totalOccurrences" : 4129, "percentContaining" : 41.29 }
   { "_id" : { "key" : "image.link" }, "value" : { "type" : "String" }, "totalOccurrences" : 4117, "percentContaining" : 41.17 }
   { "_id" : { "key" : "image.title" }, "value" : { "type" : "String" }, "totalOccurrences" : 4113, "percentContaining" : 41.13 }
   { "_id" : { "key" : "copyright" }, "value" : { "type" : "String" }, "totalOccurrences" : 930, "percentContaining" : 9.3 }
   { "_id" : { "key" : "authors" }, "value" : { "type" : "Array" }, "totalOccurrences" : 592, "percentContaining" : 5.92 }
   { "_id" : { "key" : "authors.XX.name" }, "value" : { "type" : "String" }, "totalOccurrences" : 591, "percentContaining" : 5.91 }
   { "_id" : { "key" : "authors.XX.uri" }, "value" : { "type" : "String" }, "totalOccurrences" : 307, "percentContaining" : 3.0700000000000003 }
   { "_id" : { "key" : "image.description" }, "value" : { "type" : "String" }, "totalOccurrences" : 89, "percentContaining" : 0.89 }
   

2. Collection entries:

   { "_id" : { "key" : "_id" }, "value" : { "type" : "ObjectId" }, "totalOccurrences" : 529155, "percentContaining" : 100 }
   { "_id" : { "key" : "idHash" }, "value" : { "type" : "String" }, "totalOccurrences" : 529155, "percentContaining" : 100 }
   { "_id" : { "key" : "idRaw" }, "value" : { "type" : "String" }, "totalOccurrences" : 529155, "percentContaining" : 100 }
   { "_id" : { "key" : "title" }, "value" : { "type" : "String" }, "totalOccurrences" : 529147, "percentContaining" : 99.99848815564437 }
   { "_id" : { "key" : "guid" }, "value" : { "type" : "String" }, "totalOccurrences" : 529146, "percentContaining" : 99.99829917509993 }
   { "_id" : { "key" : "link" }, "value" : { "type" : "String" }, "totalOccurrences" : 529126, "percentContaining" : 99.99451956421086 }
   { "_id" : { "key" : "fullContent" }, "value" : { "type" : "String" }, "totalOccurrences" : 529126, "percentContaining" : 99.99451956421086 }
   { "_id" : { "key" : "description" }, "value" : { "type" : "String" }, "totalOccurrences" : 505082, "percentContaining" : 95.45067135338417 }
   { "_id" : { "key" : "pubDate" }, "value" : { "type" : "Date" }, "totalOccurrences" : 494181, "percentContaining" : 93.39059443830257 }
   { "_id" : { "key" : "categories" }, "value" : { "type" : "Array" }, "totalOccurrences" : 296759, "percentContaining" : 56.08167739131257 }
   { "_id" : { "key" : "categories.XX.name" }, "value" : { "type" : "String" }, "totalOccurrences" : 296759, "percentContaining" : 56.08167739131257 }
   { "_id" : { "key" : "categories.XX.taxonomyURI" }, "value" : { "type" : "String" }, "totalOccurrences" : 50579, "percentContaining" : 9.558446957885685 }
   { "_id" : { "key" : "enclosure" }, "value" : { "type" : "Array" }, "totalOccurrences" : 32798, "percentContaining" : 6.198183896967807 }
   { "_id" : { "key" : "enclosure.XX.url" }, "value" : { "type" : "String" }, "totalOccurrences" : 32671, "percentContaining" : 6.174183367822283 }
   { "_id" : { "key" : "authors" }, "value" : { "type" : "Array" }, "totalOccurrences" : 32309, "percentContaining" : 6.1057724107303155 }
   { "_id" : { "key" : "authors.XX.name" }, "value" : { "type" : "String" }, "totalOccurrences" : 32239, "percentContaining" : 6.092543772618609 }
   { "_id" : { "key" : "enclosure.XX.type" }, "value" : { "type" : "String" }, "totalOccurrences" : 30018, "percentContaining" : 5.67281798338861 }
   { "_id" : { "key" : "enclosure.XX.length" }, "value" : { "type" : "Object" }, "totalOccurrences" : 19670, "percentContaining" : 3.7172473093894984 }
   { "_id" : { "key" : "enclosure.XX.length.floatApprox" }, "value" : { "type" : "Number" }, "totalOccurrences" : 19670, "percentContaining" : 3.7172473093894984 }
   { "_id" : { "key" : "authors.XX.uri" }, "value" : { "type" : "String" }, "totalOccurrences" : 19495, "percentContaining" : 3.684175714110232 }

A quick tutorial for running the solution. The compiled solution (jar files) can be found at the target/jar directory.

1. First run the InsertResources jar:

   java -jar InsertResources.jar

   The program accepts the following arguments:

   usage: java -jar InsertResources.jar
    -collName <arg>   the name of collection to use
    -dbName <arg>     the name of the database to use
    -filePath <arg>   the path of the file with RSS feeds
    -help             help for usage
    -host <arg>       database's host address
    -port <arg>       port on which the database is running
   

   If the user does not pass any arguments then the following default values are used:

   collName = "feeds"
   dbName = "rssdb"
   filePath = "./10K-RSS-feeds.csv"
   host = "localhost"
   port = 27017
   

2. Then run RSSDelegateWorker jar:

   java -jar RSSDelegateWorker.jar

   The program accepts the following arguments:

   usage: java -jar RSSDelegateWorker.jar
    -checkInterval <arg>   time in seconds for checking stalled feeds
    -collName <arg>        the name of collection to use
    -dbName <arg>          the name of the database to use
    -help                  help for usage
    -hostBroker <arg>      the URL of the broker
    -hostDB <arg>          database's host address
    -portDB <arg>          port on which the database is running
    -subject <arg>         name of the queue
   

   If the user does not pass any arguments then the following default values are used:

   checkInterval = 24 * 60 * 60
   collName = "feeds"
   dbName = "rssdb"
   hostBroker = "failover://tcp://localhost:61616"
   hostDB = "localhost"
   port = 27017
   subject = "RSSFEEDSQUEUE"
   

3. And finally the main worker RSSMainWorker jar:

   java -jar RSSMainWorker.jar

   The program accepts the following arguments:

   usage: java -jar RSSMainWorker.jar
    -collNameEntries <arg>   the name of collection to use for entries
    -collNameFeeds <arg>     the name of collection to use for feeds
    -dbName <arg>            the name of the database to use
    -help                    help for usage
    -hostBroker <arg>        the URL of the broker
    -hostDB <arg>            database's host address
    -portDB <arg>            port on which the database is running
    -subject <arg>           name of the queue
    -threadsNum <arg>        number of active threads
   

   If the user does not pass any arguments then the following default values are used:

   collNameEntries = "entries"
   collNameFeeds = "feeds"
   dbName = "rssdb"
   hostBroker = "failover://tcp://localhost:61616"
   hostDB = "localhost"
   portDB = 27017
   subject = "RSSFEEDSQUEUE"
   threadsNum = 10
   

   Of course one can run multiple main workers.

• Implement check for simmilarity between id's of entries of given feed using Levensthein distance.
• If similarity between id's is not found then also check for similarity between full page content using Jaccard distance.