#A Parallel in-memory database management system ##Features
- This database system is based upon the concept of shared nothing architecture among a cluster of compute nodes.
- It parallelizes the insertion of data and the processing of queries.
- Each compute node consists of two processes (A & B), process A is responsible for receiving user queries and processing final results while process B is responsible for periodically inserting the data into main memory from text file as well as processing the queries received from process A.
- At the moment, the database system is capable of working with only a fixed schema that represents a Sale Database. The schema is as follows:
| Entity | Description |
|---|---|
| sales_id | an auto-incrementing integer |
| date | yyyy/mm/dd format |
| company_id | auto-incrementing integer (starting from 1). |
| company_name | A set of unique strings. The count will match with that of company_id |
| sales_total | A value between 0.99 to 999.99 |
- It can only process a predefined set of queries on the above presented Sale Database. These queries are as follows:
-
Company Sale - Returns the total sale for each of the companies in the database. Result schema is as follows:
Entity Description company_name A unique string representing company name Total_Sale A floating point value representing the total sale for this company -
Sale by Date - Returns the Sale total for each date in the user specified date range. Result Schema is as follows:
Entity Description Date yyyy/mm/dd format Total_Sale A floating point value representing the total sale on this date -
Delete - Deletes all the sale records with date in the user specified date range. i.e delete by date range.
-
- The following are the minimum requirements to compile the program
- C11 Compiler
- OpenMP 2.0 Compiler
- In order to compile the program, go to the directory consisting of all the source files and execute the following command
- mpicc *.c *.h -o exe
- In order to run the program, please use the following command
- mpirun -np a ./exe b where a = the number of processes in the MPI environment and b = an integer between 10 and 1000 that will be used to determine the number of records to be read per signal time (which is set to 3 seconds by default).