Quantcast
Channel: SCN : Discussion List - SAP HANA Developer Center
Viewing all articles
Browse latest Browse all 6412

HANA row store performing better than column store - help me understand why?

$
0
0

I am a newbie to HANA and trying to find out the best way to use it.

We are an enterprise application company. We support ORACLE currently and exploring HANA to see if we can leverage it for better performance

 

So far what I have read is using a ROW store in HANA is not generally advised unless for special cases.

But my tests are showing otherwise.

I am doing all the tests on the HANA trial instance available on SAP Cloud

 

 

Simple test -

FCSTWIDE with a composite PK (3 columns - item,loc,startdate) followed by 100 columns (period forecast values)

FCSTWIDE -> item,loc,startdate, period,period2............period100.

The table has 1200 rows.

 

select item,loc,startdate,period1,period 2 from fcstwide takes 0.4 ms (server processing time)

If I select 10 columns from period1 to period 10, it takes 0.9 ms

Selecting 50 period columns takes 3.2 ms

selecting all 100 period columns takes 6.5 ms.

 

 

So, it seems the execution time of the queries increases kind of linearly with the number of columns selected?

Is it expected?

 

 

Now, I converted the same table to a ROW STORE.

All queries run much much faster - selecting few columns, all columns etc.

Selecting all columns from the table takes only 0.4 ms compared to 6.5 ms for the column store.

 

 

Next, I am tried putting a where clause like STARTDATE<yyyy-mm-dd

In this case also, the ROW store performed much faster (0.17 vs 0.56 ms)

 

 

After that, I tried a join with a SKU table (master data) with the FCSTWIDE based on item and loc columns.

Here also, the ROW STORE performed much faster (0.6 ms vs 8 ms)

 

 

So, in all my tests, the ROW store is way faster.

 

Can you please help me understand why I am not seeing the benefits of the column store ?

Is it because the COLUMN STORE should only be used for OLAP kind of queries and the schema representing a START SCHEMA?

Or, the performance benefit of the column store only comes into picture when data volumes are significantly higher?

 

Another question - In Hana studio, we see the server processing time and execution time followed by the time to fetch the rows.

What should I be looking into? I am thinking the server execution time is the best indicator.


Viewing all articles
Browse latest Browse all 6412

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>