Terracotta and Tokutek update data storage technologies
Updated data storage technologies rely on Flash drives and working memory to store data
By Joab Jackson | Published: 13:41, 26 September 2012
Java caching software provider Terracotta and database provider Tokutek have each released new offerings designed to entice organisations to try new ways of storing data, from running in-memory databases to using Solid State Disks (SSDs) to speed performance.
Terracotta has released a new version of its BigMemory caching software, called BigMemory Go, that can store up to 32GB of Java objects on a single server's main memory. And Tokutek has updated it flagship database so that its works more effectively with flash-based SSD storage.
Terracotta is offering BigMemory Go at no charge. Organisations can deploy the software on as many servers as they wish. BigMemory Go is a version of Terracotta's BigMemory, an in-memory database first released in 2010 that can now store up to 4TB of data in RAM across multiple servers.
Related Articles on Techworld
Unlike BigMemory itself, which was recently renamed BigMemory Max, a single BigMemory Go data cache cannot run across multiple servers. The company has released this version to provide existing users of the company's Ehcache open source enterprise Java caching software with an easy migration path to BigMemory, according to Gary Nakamura, general manager of Terracotta.
Ehcache, which Terracotta has estimated to have over 500,000 users, has a limit on the amount of material of that can be stored in memory, due to the limits of Java itself. Most Java applications have a limit of about 4GB to 6GB that can be stored in working memory without significant tuning. BigMemory Go will give Java-based enterprise applications much more room to store data in memory. Terracotta is betting that, over time, enterprise applications will grow so large that they will need to store data across the memory of multiple servers and will need BigMemory Max.
"We're bringing in-memory to the masses," Nakamura said.
BigMemory Go comes with an API (Application Programming Interface) that can be used to search through the contents of the cache. It also provides a console for management as well as the ability to write the contents of the cache to disk, for persistent storage and quick restarts.
While Terracotta encourages organisations make more use of working memory, Tokutek has updated its flagship TokuDB database to work more effectively with SSDs. TokuDB, can be used as a database engine for either MySQL or MariaDB, uses a novel indexing technology, called Fractal Tree indexing, that can speed read and write times, according to the company.
Fractal Tree can also lessen the wear and tear on SSDs, the company has claimed. This factor might be an important for organisations considering the use of SSDs in the data centre. Although they can outperform standard disk drives, SSDs can wear out more quickly if subjected to intense writing and rewriting of data.
Thanks to how Fractal Tree aggregates and rebalances data, TokuDB writes to flash memory in much larger blocks than a typical database engine. This approach reduces write amplification, in which the same data is rewritten to disk multiple times.
This latest release of TokuDB comes with a number of additional improvements that have given the company confidence in pitching the database as being optimised for SSDs. These updates include refinement of the locking and latching mechanisms, better parallelism on reads, and performance improvements in point and range queries across both primary and secondary indexes, according to Lawrence Schwartz, TokuDB vice president of marketing.
TokuDB can be licensed for $2,500 per 100GB, per year. It is available for up to 50GB for free evaluation and proof of concept projects.