Questions about deletes and partitions

[Former Member]

Hi:

We're looking at high volume time-based record management and looking at comparing regular record deletes versus using partitions (and dropping them). 

What kind of performance should we expect with doing high-volume deletes in HANA One SPS5? 

Would long running deletes block other db operations like selects/inserts/updates?

Should we be able to do dirty reads during deletions?

With respect to partitioning, HANA supports multi-level partitioning so if the the answers to the above is that deletes are blocking then we would be considering using partitioning (and dropping them). 

With respect to columnar databases, is using partitioning a recommended way to go (best practice) in HANA?

Thanks

Tipton

[danielculp]

Hi,

I would opt for paritioning as well. it should be more efficient and faster than a DELETE WHERE.

Partitioning is common practice in HANA clusters with OLAP like workload. In this case you spit the tables accross nodes to make use of inter node parallelization.

Hope this helps.

Best Regards

Daniel

[swapan_saha]

Hi Tipton,

I see you have at least two points in your message. First part is related to expected performance for your use case and second part if there is any recommendation for your use case.

Let me respond to the first part of the message.  For a similar number and type of CPU cores, memory and I/O configuration, SAP HANA One is comparable to SAP HANA on-premise. Your actual performance will vary based on the nature of workload.

Thanks,

Swapan