There is no support for pagination with Column Repeat layout .
You can try achieving it in a custom way with the following steps if the source is a pagelist (If RD making a db query for every page using RetrieveReportMetaData activity passing correct set of paging parameters is not perform-ant)
1)Maintain two pagelists on the clipboard :
-one with total set of records
-other with current set of records that can be viewed in a page (onload populate this page list with first page set of records).
2)Bind the column repeat to 2nd pagelist property
3)Configure buttons for navigation
- set Refresh this section event Onclick of these buttons and run a pre-activity which does the clearing earlier set of records and populating current set of records (next or previous page set) taking relevant parameters like page size , current page index etc..
This activity copies the correct set of records from 1st pagelist to the 2nd pagelist which is a source to column layout and this will be reflected in UI as well since refresh is done.At any point of time in clipboard the 2nd pagelist should hold only the set of records in current page.
This looks like an very interesting take towards pagination. But the only caveat I see is that we have to maintain two sets of clipboard pages (1 with whole set of data and the other with current page data). If the clipboard page holding the entire data is huge then the clipboard foot print will be compromised. Instead I suggest a small optimization, configure the pagelist bound to column repeat layout with "Refer to data page" setting (in property rule form). This data page has to be a parameterized one (parameter could be the page index itself), so that whenever we request for next page, data page will be populated again with the next set of data and the pagelist will be implicitly populated.
The main benefit we get with this approach is that once the user has paginated till the last page, there is no more DB hits as the data page is already populated with all sets of data with different parameters. Also this will be efficient in terms of both memory and performance.