We’ve researched a cloud database solution for a couple of months; the project name is Cloudbean. The project aims to bring a specially structured database solution which is scalable, high-performance distributed database system. The optimization process is now going on to make the solution faster than before. And now we unveil the result of performance test conducted recently. The Cloudbean version is not the newest that the upgrade and code migration are still in progress. The tested Cloudbean version is 1.2, and the test environment – hardware specification – looks like below.
- Intel Dual-core E6500 2.93GHz
- Samsung 2GB Memory
- WD SATA2 HDD 500GB (7,200 RPM)
The test is done with 100,000 key-value pairs in a single node. Each key and value sizes are 8 bytes. The following is the result of the test in reading and writing comparing to other noSQL solutions like membase and Cassandra.
- membase: writes data for about 28 secs, reads for about 12 secs in average
- cassandra: writes data for about 30 secs, reads for about 72 secs in average
- cloudbean data service: writes data for about 26 secs, reads for 2 secs in average
The optimization in this project means to minimize the data writing to disk. We have developed the prototype version 1.2, which is not stable yet though, the next version 2.0 would be more stable, satisfying ACID properties. Cloudbean is designed to offer a special purposed data model so that we could be prepping for a new kind of cloud service. Long way to go, but we have a vision for delivery.
