This project consists of simulating how the priority-based support provided by NVMe devices can be exploited to enhance the Quality-of-Service in Database Management Systems.
Our proposal simulate the NVMe Weighted-Round-Robin queues of an NVMe device, as well as we simulate traces from the different database threads.
- Cassiano Campes
- Dmitry Lachinov
-i <input path> -o <output path> -w <weights path> -rt <read time> - wt <write time> -qs <dev queue size>
-i path to trace csv
-o path to output file. File either shouldn't exist or be modifiable
-w path to weights csv
-rt page read time (in units)
-wt page write time (in units)
-qs size of the device queue
-i C:\Users\Dmitry\Documents\GitHub\DBProject\trace.csv -o C:\Users\Dmitry\Documents\GitHub\DBProject\output.csv -w C:\Users\Dmitry\Documents\GitHub\DBProject\weights.csv -rt 10 -wt 100 -qs 10
-tx <transaction size> -cp <checkpoint size> -cmp <compaction size> -file <output file> - req <number of requests> -trigger <compact trigger>
-tx transaction buffer size (in number of pages)
-cp maximum checkpint size per operation
-cmp maximum compaction size per operation
-file filename to store the simulated traces
-req number of requisitions (traces) to be generated
-trigger trigger value (in # of pages) that starts a compaction service
~/Develop/DBProject$ ./tracing -tx 100 -cp 5 -cmp 10 -file "parse2.csv" -req 1000000 -trigger 1000