If you execute the statement SELECT column 1, column 2 from table where 'condition'; then it will partially load the table to the memory.
You can check this in the m_cs_tables in main memory size and loaded column.
For the partially loaded memory, the loaded column will contain value as PARTIAL.
Regards,
Ravi
what you say is CORRECT , but would prefer a more elegant way to doing it rather than running a actual select query.
The idea is suppose we load data for the last 10 years at the begining and then after doing some initial sampling and reports , we only use the current and past 4 years of historical data .
Just so we dont run out of RAM space , we dont want to load the first 5 years , but just the last 5 years into the RAM but nor do we want to delete them from the persistence storage area .
That is exactly what SAP will come up, in very near future, with the option of HOT and COLD storage. With that you can define the historical data as Cold data which is available in the cold storage (Persistent storage) and recent data in Hot main memory.
Regards,
Ravi
Another alternative to that is the NLS (Near Line Storage). Currently Sybase NLS solution is supported with HANA.
Regards,
Ravi
Hello Rajarshi,
why would you want to partially unload data from RAM?
Do you do that on other DBMS as well? Removing cached blocks from the db cache?
Also, be aware that once you start modifying data in any column store table and eventually a delta merge needs to be performed, the unloaded data needs to be re-loaded to memory again.
Apart from that, HANA already employs a automatic unloading of less used data in case of memory shortage.
regards,
Lars
Lars : This is a hypothetical situation where the HANA has low RAM, and thus needs to unload certain tables which are not needed .
My company ( Accenture ) is often being asked this question during our proposals phase .
Lars
Do you have any information on SAP's algorithms for automatic unloading?
I don't have any specifics about the algorithm to share, but afaik it's a straight forward least-recently-used algorithm ("what hadn't been touched for the longest time gets kicked first").
cheers, Lars
