Problem:
We generated the Vertica compression output using "collect_diag_dump.sh -c". However there are some projections for which the compressed size (Vertica (MB)) is greater than raw size.
Please see output below. I have highlighted the relevant columns for easy viewing.
TableName PjCnt PjType RowCount RawSize(MB) Vertica(MB) Compr.Ratio
=============================================================================================================================================^[[0m
temp.dwh_fact_invites_rtm_guid 3 S 145985594 46809.04 53595.23 0.87
temp.dwh_fact_invites_client_Type 3 S 145985594 46809.60 52706.82 0.89
temp.dwh_fact_day_id 3 S 145985594 46808.90 52610.92 0.89
temp.dwh_fact_invites_suc_ct 3 S 145985594 46810.16 51368.85 0.91
temp.dwh_fact_mult 3 S 145985594 46808.35 48434.29 0.97
temp.dwh_fact_invites_rtm_guid 3 S 145985594 46809.04 53595.23 0.87
temp.dwh_fact_invites_client_Type 3 S 145985594 46809.60 52706.82 0.89
temp.dwh_fact_day_id 3 S 145985594 46808.90 52610.92 0.89
temp.dwh_fact_invites_suc_ct 3 S 145985594 46810.16 51368.85 0.91
temp.dwh_fact_mult 3 S 145985594 46808.35 48434.29 0.97
Solution:
The column PjCnt in output above means projection count for corresponding tables. In this particular case, each table had 3 projections (1 super and 2 buddies).
The compression ratio is between Raw Size of the table and Vertica size of all projections on that table. Raw size calculated only one time where as Vertica size is the sum of all projections storage size. As there are 3 projection on those table, Vertica size on those tables increased slight more than raw size and compression ratio is less than 1.
This is expected. If we keep adding more projection on a table, Vertica size grows where as Raw size remains same and compress ration decreases.
