HANA tables size on disk is reducing

[nicholas_chang]

Hi Lars,

Came across your statement "The primary storage is memory for SAP HANA, so whatever ends up on disk is equivalent to a database backup." here and i got one question:

Any clue what cause db bakup size (sum(allocatd_page_size) is much greater than disk size (sum(disk_size) for M_TABLE_PERSISTENC_STATISTICS)? The size diff is around 50GB~~~ Delta merge is running correctly.

this is quite abnormal with such a large gap. Checked on others systems, where the gap is around 5-10GB.

Hope to hear from you soon,

Thanks in advance,

Nicholas Chang

[lbreddemann]

First off, repeat after me: I shall not hijack/resurrect other threads!!

Not sure where you get the allocated_page_size from!? M_CONVERTER_STATISTICS I guess...

If so then the answer has to be: not everything in SAP HANA is stored in tables. E.g. undo information would not show up in the TABLE_PERSISTENCE view.

Martin has written some really nice FAQ notes, including one on storage and I/O. I propose you check them out and see if maybe you are looking at some form of delayed garbage collection.

- Lars

[nicholas_chang]

Hi Lars,

"I shall not hijack/resurrect other threads!!"

Thanks for answering btw 😉

Yup, i've went thru those FAQ, admin guide, training guide and etc several times but still greasing my elbow to get the answer:

I did a comparison with other HANA system, as you can see, the Disk Size in first row was 40GB smaller compare to the backup size and used data volume size; whereas , value of other systems are pretty close.

Any clue what causing this that savepoint is write and updated to the used data volume (M_CONVERTER_STATISTICS) but not disk size (M_TABLE_PERSISTENCE_STATISTICS) that caused such a hugh difference?

What other info that persisted in memory and eventually to data volume but not to global disk?

Fyi, below are running fine:

i) delta merge

ii) savepoint

iii) mvcc

iv) total RS+CS tables is 2++ greater than M_TABLE_PERSISTENCE_STATISTICS. (which is true as this apply to other systems)

v) temporary table is small in size.

vi) couldn't find any note relate to this bug runing on 85.02.

i believe internal column will be written into data volume as well as global_disk?

Thanks,

Nicholas chang

[lbreddemann]

Hmm... I'd guess there might be some hick-up in the M_TABLE_PERSISTENCE stats.

You may want to have this checked by SAP support.

- Lars

[nicholas_chang]

alright. Thanks Lars!