Apache Arrow Reviews & Pricing

Key Features

• Libraries: Work with data through its native libraries for C++, C#, Go, Java, JavaScript, MATLAB, Python, R and Ruby — no need to implement a columnar format for each project separately. 
• Faster CSV Reading: Reads CSV into Pandas more than 10 times faster by virtue of its columnar storage design. Stores and reads data in parallel through record batches — 2D data structures containing columns of data of equal length. 
• Faster UDFs in PySpark: Efficiently transfers data between Java Virtual Machines (JVMs) and Python processes with vectorized user-defined functions (UDFs), doing away with serialization/deserialization and enabling faster data processing. 
• Record Batches: Reads a folder containing many data files, and even subfolders, into a single dataframe by virtue of record batches. Some libraries like C++, Python and R support reading entire directories of files and treating them as a single dataset. 
• Read/Write Parquet Files: Reads parquet files into Python and R through translators that convert data into language-specific in-memory formats. Writes data held in-memory in tools like Pandas and R to disk in Parquet format. 
• Memory Mapping: Works with data bigger than allocated memory space and allows data sharing across languages and processes through local mapping of its IPC files. 

Limitations

•  Its memory layout isn’t ideal for workloads involving accessing multiple attributes from a single entity, as in OLTP loads. 
•  Its code can be slow to execute sometimes. 

Suite Support