The C++ API is for those who need the speed of a native implementation for mapreduce algorithms. You can use the API to write mapreduce jobs in C++ and send them to the Yothalot cluster.
With the C++ API you write an executable (so you have to write a main() function too). This executable has to be installed on each of the servers in the Yothalot cluster. The Yothalot process will start up this executable with specific command line arguments and input data, so that (a part of the) mapreduce algorithm gets executed.
Besides the main() function that you have to supply, you also have to implement the map/reduce algorithm. For this, the C++ offers a virtual base class that you have to extend in your program. This base class has various methods for mapping, reducing and writing that should all be implemented by you.
If you want to use the C++ API you only have to install Yothalot. Make sure that Yothalot is also installed on the machine where you build your mapreduce programs since building the program requires the yothalot library that comes with the Yothalot installation.
You can run a mapreduce task on the Yothalot cluster by extending a virtual base class that is supplied by the Yothalot framework:
Creating your program
To turn your algorithm into an executable that can run on the Yothalot cluster, you have to add a "main()" function to it, and process the command line arguments that are passed to it by Yothalot. The C++ API offers a very useful utility class for parsing these arguments, so that your main() function normally does not need more than 3 lines of code:
It is also possible to use Yothalot to run programs that do not hold a mapreduce algorithm. The requirements of these programs and there use is discussed in:
Starting a Yothalot job
You start up a job by calling the Yothalot program and the appropriate command line arguments. The options are discussed on this page:
The classes mentioned above are sufficient for writing and running mapreduce jobs. As an extra service however, the Yothalot C++ API also allows comes with a couple of extra utility classes.
The Yothalot::Fullname class can be used for converting absolute pathnames into pathnames that are relative to the GlusterFS mount point, and the other way around. This is especial useful if you use different mount points on different servers.
Full class reference
The following classes are offered by the Yothalot framework:
- Class Yothalot::Input
- Class Yothalot::Inputs
- Class Yothalot::Fullname
- Class Yothalot::Key
- Class Yothalot::MapReduce
- Class Yothalot::Output
- Class Yothalot::Record
- Class Yothalot::Reducer
- Class Yothalot::Result
- Class Yothalot::Scalar
- Class Yothalot::Tuple
- Class Yothalot::Value
- Class Yothalot::Values
- Class Yothalot::Writer