Let´s talk about disadvantages...

[jgleichmann]

Hi HANA-fans,

everyone talks about the performance and the advantages of HANA - I want to talk about the disadvantages

With my dangerous superficial knowledge I found this disadvantages about HANA. I hope someone can write down a few statements to the following points:

1) currently only one big maschine - which is not really perfect for HA scenarios ( in some documentation you can read about a distributed system - like BWA) - no docu tells me the truth about the HW architecture

2) no PITR (point-in-time recovery - user errors -> deleting data accidentally or something like that)

3) HA scenarios are not covered by SAP - vendors will be responsible (standby host(s)?)

4) no migration back to standard RDBMS, so SAP take your flexibility away ( no R3load export possible?)

5) System copies - how they can be performed? (more interesting later for ERP systems)

I want to know if some of these topics are implemented in the next releases or if they planned at all.

Regards,

Jens

[former_member93896]

Hello Jens,

just a few comments:

1) currently only one big maschine

Not anymore. Please see http://service.sap.com/pam

2) no PITR (point-in-time recovery)

Planned to be part of HANA 1.0 SPS 3

3) HA scenarios are not covered by SAP

Planned to be part of HANA 1.0 SPS 3. The solution still depends on the HW vendor since for example different file systems are used.

4) no migration back to standard RDBMS, so SAP take your flexibility away

Why would you?

5) System copies

Just a matter of time until this will be possible.

Regards,

Marc

SAP Customer Solution Adoption (CSA)

[jgleichmann]

Hi Marc,

at first a big thanks to you for your statements!

1) currently only one big maschine

Not anymore. Please see http://service.sap.com/pam

-> So if I understand the PAM right, currently only IBM supports multiple hosts all other vendors only supporting single server. If you are using multiple hosts I think it´s really difficult to sync all this hosts in time, so that you don´t have poor performance. This is realized via InfiniBand technology, right? I heard some issues regarding multiple hosts running with HANA...

4) no migration back to standard RDBMS, so SAP take your flexibility away

Why would you?

-> hehe, sure I think always forward and so I have also a look on other in memory RDBMS which may be support by SAP. I this planned? I hope so 😄

@Vitaliy: I hope HP is also working on a multiple host solution for HANA

Cheers!

Jens

[Former Member]

For me it's:

1. Data Lifecycle Management - I'm paying (A LOT mind you) for data storage, so why report on old data that I don't use? Is there any time partitioning.

2. Running out of memory - What if the processing takes up too much memory? What if my data grows too quickly and I don't have enough memory left? What then?

3. I'm still a bit confused about deletes. As far as I know, everything is "append-only". How does selective deletion work?

4. HA is probably the biggest unknown. Apparently there are solutions from HW vendors, but this is still a big unknown for me.

Good topic Jens!

[former_member93896]

Hello Mike,

1) Agree. It does not make much sense to keep all historical data in memory if it isn't accessed regularly. We plan to include a data aging mechanism in a future HANA version (maybe disk-only tables or NLS).

2) If memory if full, the system will start swapping (i.e. unloading table columns). Since this has a significant impact on performance, you don't want to run into this situation. The system is simply not sized properly.

3) It's a SQL database. You can do DELETE FROM table WHERE ... on any table. What happens inside HANA is a different story and depends on the type of table (ROW, COLUMN SIMPLE, COLUMN HISTORY). In simple terms, column tables have two memory structures: main and delta. Changes are indeed appended to the delta part, which is then merged into the main part asynchronously.

4) HA: Significant enhancements are planned for HANA 1.0 SPS 3 (see also PAM).

Regards,

Marc

SAP Customer Solution Adoption (CSA)

[Former Member]

Hello Jens,

for me the main disadvantage of such a new software like HANA is that it is brand new. So the early adopters have to anticipate glitches or maybe even data loss (keeping fingers crossed). I am really eager to see how HANA is doing as a database for BW, but I wouldn't use it for productive BW systems right from the start. The good point of SAP BW is that your source systems can theoretically deliver data simultaneously to a conventional BW/RDBMS instance and BW/HANA, so in the early BW/HANA days it should be possible to gain experience without any risk.

Maybe I am too optimistic about it, but I don't think your five points apply to HANA 1.0 SP3. For example, why would SAP write an interface to import data into HANA via R3load, but not export from HANA via R3load? This doesn't make sense to me. Just think about a scenario where you export from HANA via R3load in order to import into another HANA via R3load. Such scenarios should be possible even though it is "just" a homogenous migration. Of course, as Vitaliy wrote we'll know better soon.

Regards,

Mark

[Vitaliy-R]

Jens, as you said as well - let's wait for Service Pack Stack 3 of HANA 1.0, that should be released in just 2 weeks from now, and then return to these questions. I am curious myself. The good thing I am hearing is that there is a hope to get documentation for SPS3 released the same day as the code, not with delay as it happened with SPS2 in June. Regards. -Vitaliy

[jgleichmann]

Hello HANA testers,

in consideration of the fact that there are a lot of posts regarding installation errors I guess HANA 1.5 was already released.

So I want to push this topic a little bit. Looking foward to here your feedback!

Thanks in advance!

Cheers,

Jens