cancel
Showing results for 
Search instead for 
Did you mean: 

7.6.5.x: UTF-16 vs. UTF-8/CESU-8 - how to...

Former Member
0 Kudos

As just attended the MaxDB-Infoday yesterday I´m curious about the possibility to compress (?) the non-key non-index fields in certain tables.

Is there any information available how to do that? Or are we talking here about USEUNICODECOLUMNCOMPRESSION parameter?

Markus

Accepted Solutions (0)

Answers (1)

Answers (1)

roland_mallmann
Advisor
Advisor
0 Kudos

Hi Markus,

yes, that's the USEUNICODECOLUMNCOMPRESSION, i.e. non-key char columns can be compressed.

A full UTF8 implementation is in the works too, but that will take more time.

Regards,

Roland

Edit: you can also 'compress' unicode (UCS2) tables using the ALTER TABLE <table> PACKED command

markus_doehr2
Active Contributor
0 Kudos

> yes, that's the USEUNICODECOLUMNCOMPRESSION, i.e. non-key char columns can be compressed.

Interesting... because if I do an R3load export/import of a 2,1 TB system, its size decreases only some 100 MB, but not more...

> A full UTF8 implementation is in the works too, but that will take more time.

yes - I heard that...

> Edit: you can also 'compress' unicode (UCS2) tables using the ALTER TABLE <table> PACKED command

Ok... I'll try that with some bigger tables

Markus

roland_mallmann
Advisor
Advisor
0 Kudos

Hm, only 100MB sounds a bit low, VERY low I'd say, but it really depends on how many and how big/wide the non-key char columns of your system are.

markus_doehr2
Active Contributor
0 Kudos

> Hm, only 100MB sounds a bit low, VERY low I'd say, but it really depends on how many and how big/wide the non-key char columns of your system are

Well - YES - that's why I was asking...

The system was initially installed with 2.2D and then upgraded through various releases till ERP 6.0. We have double-byte characters in the system (chinese, korean) but the main data (I'd say roughly 90 %) are ASCII.

I imported the same export into an Oracle system (to get a comparison) , it used finally 1,1 TB (so almost 50 % less).

So I think we have to wait until the UTF-8/CESU-8 is fully implemented including the B*trees...

Markus

roland_mallmann
Advisor
Advisor
0 Kudos

Just a small additional remark: indexes themselves are not compressed in UTF8 with the columncompression, but the columns from the basetable are, as long as they are not part of the key and if they are char columns.