Benchmarking Top NoSQL Databases - DataStax（基准测试前DataStax NoSQL数据库）.pdfVIP

Benchmarking Top NoSQL Databases Apache Cassandra, Couchbase, HBase, and MongoDB Originally Published: April 13, 2015 Revised: May 27, 2015 / Table of Contents Executive Summary3 Overview3 Test Methodology and Configuration 3 Cassandra Software 4 Couchbase Software4 HBase Software 5 MongoDB Software 5 Workload Selection 5 Result Collection and Aggregation6 Throughput by Workload 7 Load process 7 Read-mostly Workload8 Balanced Read/Write Mix 9 Read-Modify-Write Workload 10 Mixed Operational and Analytical Workload 11 Insert-mostly Workload 12 Average Latency by Workload 13 Read-mostly Workload13 Balanced Read/Write Mix Workload 14 Read-Modify-Write Workload 15 Mixed Operational and Analytical Workload 16 Insert-mostly Workload 17 Database Idiosyncrasies and Difficulties Encountered17 Configuration Details 18 Cassandra18 Couchbase 18 MongoDB 19 HBase 19 Page: 2 End Point Corporation 304 Park Avenue South, Ste 214 New York, NY 10010 +1 212-929-6923 Executive Summary End Point performed a series of performance tests on Apache Cassandra, HBase, and MongoDB using the YCSB benchmark in February 2013. Because performance enhancements have been made to each database since that time, End Point was commissioned by DataStax to conduct a retest of each database, with Couchbase also included as it is sometimes used as a NoSQL caching layer for certain application workloads. The overall conclusions of the performance tests closely mirrored the results of the previous benchmark, with Cassandra outperforming the other NoSQL databases in both throughput and latency. Depending on the test, HBase or Couchbase delivered the next best result, with MongoDB