The new version brings better API, new features, performance improvements, enhanced UI, and stability to Tachyon. Major changes are documented below:
Downloads for the release can be found here: Downloads.
Java-like file API
- Tachyon has an API similar to JAVA file with InputStream, OutputStream API.
- Efficient support for memory mapped IO.
Native support for raw tables
- Table data with hundreds or more columns is common in data warehouses. Tachyon provides native support for multi-columned data.
- The user can choose only to put often queried columns in memory.
- Hadoop MapReduce and Spark can run with Tachyon without any modification.
Pluggable underlayer file system
- Tachyon checkpoints in memory data into the underlayer file system. Tachyon has a generic interface to make plugging in an underlayer file system easy.
- Tachyon now supports HDFS, and single node local file system.
- The new web UI is more user friendly.
- User can browse the file system easily.
- With debug mode, admins can also view detailed information of each file, including locations, checkpoint path, etc.
Command line interaction
- Users can use ./bin/tachyon tfs to interact with the file system.
- It is simple to copy data in and out of the file system.
- We have significantly simplified the deployment process.