Skip to Content
0

Partition on Single-host HANA System

Jul 31, 2017 at 05:18 AM

315

avatar image

Hi Experts

Partitioning is best strategy when you anticipate the good data size growth of your table and chances of reaching the 2 billion threshold value of your table.

Partitioning helps to distribute the data from table across your partitions which are spread at multiple nodes that contribute to HANA System in multi-host system.

What happens when I have single host system with Range partitioning of data (Partitioning on quarterly based data) , will HANA System create virtual partitions, and mimic the way partitions work in multi node system.

Would partition really benefit in case of single host system?

10 |10000 characters needed characters left characters exceeded
* Please Login or Register to Answer, Follow or Comment.

4 Answers

avatar image
Former Member Aug 03, 2017 at 07:48 PM
-2

Partitions totally depend on nodes availability on servers.

Lets say you have 2 nodes on server and You defined partitions as 3

Then node 1 will have 2 and Node 2 will have 1 partitions.

This is helpful when your data size is very high, Else not needed.

Only single node will not help much

Show 3 Share
10 |10000 characters needed characters left characters exceeded

What do you base your opinion about the benefits of partitioning on?

Why do you say that partitioning on a single host system wouldn't be beneficial?

It's easy to show examples where partitioning tables on a single instance is very beneficial - even for situations where the volume of data is not close to 2 billion records.

0
Former Member

What do you base your opinion about the benefits of partitioning on?

- Partitioning is definitely advantage to improve performance by using HANA core parallel processing.

Why do you say that partitioning on a single host system wouldn't be beneficial?

- Not exactly, We need to identify the need of partition approach 1.Performance 2.limit to store 2 billion records.

Now 1 node can support multiple partition which can extend the limit of 2 billion records/table. But performance wise massively parallel processing wont help as we have single host/node.

Now with multiple nodes/host both scenarios are covered.

It's easy to show examples where partitioning tables on a single instance is very beneficial - even for situations where the volume of data is not close to 2 billion records.

I agree, But not sure how this is beneficial as there will be single index server/core processor.

Also we need to think from insert/update perspective,It has to determine in which memory area record should go.

0

"But performance wise massively parallel processing wont help as we have single host/node."

A single host can have hundreds of cores nowadays. Using those in parallel is what HANA is all about. This statement of yours is as wrong as it gets with regards to HANA.

"I agree, But not sure how this is beneficial as there will be single index server/core processor."

Why would there be just a single core? There is not one single platform HANA runs on that is single core. Even my small NUC test-server is a quad core i7 that provides 8 parallel threads.

"Also we need to think from insert/update perspective,It has to determine in which memory area record should go."

Yes, when using partitioned tables, HANA needs to map data to the partitions. But this is not the part of the processing that takes a lot of time. Adding data to the delta store and running merges takes time. Allocating larger and larger continuous memory blocks takes time.

Hmm... considering your comments...

1
Lars Breddemann
Aug 03, 2017 at 01:07 PM
0

Not sure what you mean by virtual partitions... you may want to read up on HANA concepts.

Share
10 |10000 characters needed characters left characters exceeded
avatar image
Former Member Aug 10, 2017 at 07:28 AM
0

Hi Anirudhdha,

Yes, you are right. A column store table or table partition can hold maximum 2 billion data records with 314,xxx,xxx,xxx as maximum integer value for contiguous memory area. So even if you have a big single host system where you can not scale out your Hana database solution then also table partitioning will help you to store huge data records (exceeding 2 billion limit) in a single table.

Hope this answers your query.

Best Regards,

Sourav Brahma

Share
10 |10000 characters needed characters left characters exceeded
avatar image
Former Member Aug 15, 2017 at 08:55 PM
0

Hi Aniruddha,

In case you have big table in HANA even though it may be less than 2 Billion records, Partitioning can always help for performance impovement and also save memory in case when HANA is going to load data into RAM.

You can select which partition should be loaded at startup, Queries execution will also be faster due to less scanning records.

Thanks,

Shivam

Share
10 |10000 characters needed characters left characters exceeded