MongoDB and Apache HBase: Benchmarking
Download as PPTX, PDF3 likes4,985 views
This document compares MongoDB and HBase on technical features and performance. MongoDB is a document-oriented database using JSON-like documents with dynamic schemas, ad-hoc queries, and indexing. HBase is a column-oriented NoSQL database modeled after Google Bigtable and stored on HDFS. The document benchmarks the two databases on loading performance using YCSB and shows MongoDB generally performs better, especially for read-heavy workloads. Popular use cases of each are also listed.
MongoDB and Apache HBase: Benchmarking
- 1. AND Products comparison
- 2. Technical overview Programming language C++ Java Language Bindings & Clients C, C++, Erlang, Haskell, Java, JavaScript, .NE Java, Jython ,Groovy DSL , Scala, REST T (C# F#, PowerShell, etc), Perl, PHP, Python, Rub y, Scala. Protocols Mongo Wire Protocol Apache Avro, Thrift, REST First public release and current state Feb 2009 Last release 2.0.2 14th Jul 2010 Last release 0.92.0 23th January December 2011 2012
- 3. Technical overview Querying Mongo Query Language Filter Language Atomicity Conditional + Consistency + + Isolation - + Durability + - Periodic-Update Secondary Index Indexing of embedded element, Filter Query Secondary Indexes compound key Dual-Write Secondary Index Summary Tables Map/Reduce Supports Sharding + + Replication + + Revision control - +
- 4. MongoDB features Document-oriented Capped Collections Greed FS Indexing Map Reduce Query language JSON/BSON Eventually-consistence
- 5. HBase features • Column oriented(after Google big table) • Bloom filters on per column basis • MapReduce • Secondary Indexes • HDFS based • Revision controll
- 6. MongoDB configuration example
- 9. MongoDB use cases Git Hub : the social coding site, is using MongoDB for an internal reporting application. РосГос затраты: RosSpending is the first Russian public spending monitoring project.. Disney: common set of tools and APIs for all games within the Interactive Media Group, using MongoDB as a common object repository to persist state information. Over 300 of companies have prodact deployments of mongoDB
- 10. HBase use cases Facebook : Real-Time messaging Over 152 billions messages monthly Adobe: 30 nodes social services ,data and processing for internal use. Explorys: over a billion anonymized clinical records Mozilla Socorro : Crash reporting system Powered by about 40 companies
- 11. Benchmarking • Enveroment: Amazon Elastic compute cloud. • Testing tool – Yahoo Cloud Service benchmark(YCSB)
- 12. 2000.00 4000.00 6000.00 8000.00 10000.00 12000.00 0.00 0 1200 2400 3600 4800 6000 7200 8400 9600 10800 12000 13200 14400 15600 16800 18000 19200 20400 12hours of loading. 21600 22800 24000 25200 167.600.000 for 4 shards 26400 27600 28800 30000 31200 32400 95.000.000 records for 2 shards 33600 34800 36000 37200 38400 MongoDB Benchmarking. 39600 40800 2 shards loading 4 shards loading 42000 43200 44400
- 13. 50% reads 50% updates 12000 10000 8000 2 shards update 2 shards read 6000 4 shards update 4 shards read 4000 2000 0 500 1000 2000 3000 3300 5000
- 14. 95% reads 5% updates 16000 14000 12000 10000 2 shards update 8000 2 shards read 4 shards update 6000 4 shards read 4000 2000 0 500 1000 2000 3000 4000 5000
- 15. Read only performance 10000 9000 8000 7000 6000 5000 2 shards 4000 4 shards 3000 2000 1000 0 500 1000 2000 3000 4000 5000
- 16. Read Insert performance 250000 200000 150000 2 shards insert 2 shards read 100000 4 shards insert 4 shards read 50000 0 200 300 400
Editor's Notes
- #13: DataNodes are constantly reporting to the NameNode. Blocks are stored on the Data Nodes.