Is TPC-D suitable for parallel DBMS evaluation?
-
No scalability performance metrics suitable for parallel data query evaluation i.e., query time vs. number of instances (CPUs), and/or, query time vs. available I/O resources ( I/O channels, drivers, disks and disk partitions)
|
|
Scalability of a parallel database system relies on both CPU resources and I/O resources.
|
|
By using multiple instances (CPUs), parallel server/query decreases query response time and increases throughput.
|
|
By using partitioned data on multiple disks (with multiple I/O channels), parallel server/query achieves higher I/O bandwidth for large DBs.
|
|
The two characteristics are equally important and must be benchmarked together. In current TPC-D specification, neither performance metrics is defined.
|
