Mohammed Shahnawaz

> You choose a segmentby column that is a unique column which will lead to very bad compression rate. > > ``` > sven@dev[709352]=# SELECT count(*), count(distinct id) from c_usge...

Current scope of SkipScan : ========== **_Hypertable_** **Query#1**: explain select distinct sensor_id from sensor_data; ```explain select distinct sensor_id from sensor_data; QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------------------- ------------------- Unique (cost=6.23..11859.26 rows=200 width=4) -> Merge...

Implementation Details: ============ **Planner Changes:** - [x] Transform **CustomScan(DecompressChunk)->IndexScan/IndexOnlyScan** INTO **CustomScan(DecompressChunk)->CustomScan(SkipScan)->IndexScan/IndexOnlyScan.** - [x] Modify get_distinct_var() to map var for compressed chunk

Hi @ebreijo, Thanks for bringing this up. Help us with the following information. - Whether this issue showed up from the very first cycle of compression job after the upgrade...

Hi @ebreijo , Thanks for being patient on this. We are going through this issue in detail and trying to find a trade-off pattern between Seq Access and Random Access....

@ebreijo , It would be interesting to compare the I/O stats (cache hits and misses) between Index path and Non-Index path compression. - Please identify Two uncompressed chunk, suppose Chunk1...

@ebreijo , Thanks again for highlighting this issue. **Brief History about Indexscan path introduction** ----------------------------------------------------------- Before 2.11 cost of compression can be expressed as cost(Seq_IO) + cost(Sort). Usually, cost of...