In order to ease the development of the project and to allow students to develop faster solutions to the project, we have updated the parameters of the submit command.

The parameters of the client submit command will be:

• The path to an input file, stored on the local file system, which will be used as input

for the job. Note that the input file is accessible only on the local node on which

the application and its client are running. The input file, instead, is not accessible by

worker nodes, which need to retrieve the input data by the client.

• The number S of splits in which the input file shall be partitioned.

• The path of an output directory on the local file system, which will be used to store

the files containing the output of the processing of the input splits, and which shall be

named“1.out”, “2.out”, ..., “S.out”.

• A class implementing the IMapper interface, which exposes a single method, i.e.:

namespace PADIMapNoReduce {

      public interface IMapper {

            IList<KeyValuePair<string, string>> Map(string fileLine);

      }

}

• The path to the class library (dll) where the IMapper implementation is.

The parameters to the PuppetMaster's SUBMIT command were also updated and are now:

SUBMIT <ENTRY-URL> <FILE> <OUTPUT> <S> <MAP> <DLL>: Creates an application on the

local node. The application submits a job to the PADIMapNoReduce platform by system by contacting the worker at <ENTRY-URL>. The job is defined by the following parameters:

– <FILE> is the path to the input file. The file will be subdivided into <S> splits across the machines in W.

– <OUTPUT> is the path to an output directory on the local filesystem of the application, which will store one output file for each split of the input file name “S1.out”, “S2.out”, ..., “S.out”.

– <S>, i.e. the number of splits of the input file, which corresponds to the total number of worker tasks to be executed.

– <MAP> is the name of the class implementing the IMapper interface.

– <DLL> is the path to the class library (dll) where the implementation of IMapper is stored.

Remember that the IMapper implementation MUST be independent from the PADIMapNoReduce code.

Here you can find an example of how you can pass any class library to a .Net Remoting server and, on the server side, dynamically load the assembly, instantiate a class and call the instantiated object's Map method. As you can see in this example, the LibMapper DLL containing the Mapper is not referenced by any of the other projects and therefore may only be known at run-time.

• Exemplo 1
• Exemplo 2
• Exemplo 3