The second validation environment consisted of a medium size cluster with low memory/ vCPUs node configuration:
Figure 8. Validation Environment 2 Results
Figure 9. 3 Workers vs 7 Workers 128GB/ 32vCPUs
The following observations were made:
- Three (3) iterations of 99 TPC-DS queries executed successfully for 1 and 5 concurrent users.
- Average execution time for 99 queries with non-partitioned data is 1.75 times that of partitioned data for both 1 and 5 concurrent users.
- With 10 concurrent users, some of the queries failed due to lack of sufficient memory. This indicates that as concurrent workload increases, we need to increase the hardware capacity – either by increasing the number of worker nodes or by having larger size of memory/ vCPUs per worker node.
- The variations in the Average execution time per query across the 99 TPC-DS queries is due to the difference in the complexity of the queries and the volume of data being analyzed.
- Compared to the smaller cluster size of three (3) worker nodes, the query response times improved with the bigger cluster size of seven (7) worker nodes.