发信人: xlxie (三老爷), 信区: Database
标 题: Re: 请教sql server DB 大侠
发信站: BBS 未名空间站 (Thu Jan 25 22:21:38 2007)
If your DB is only used to record inserting data and seldom do delete or
Then, you can record update/insert date as a seperate column.
Use "select into" to do incremetal load and
if your DB is big, you will have to think about table partition.
For OLAP reporting, do you need the data warehouse and mart at the same
or you need a data warehousing server and a mart server?
different approach will lead to different design.
restore backups is not a quite efficient way.
If you have to do it, do a full back up weekly and incremental/differential
If you want to do OLTP => OLAP, that is not the solution.
Probably you will endup using DTS/SSIS to transfer the data between your
Production server and reporting servers.
Some coding will need to be done to make that happen.
【 在 mirthcyy (百无聊赖) 的大作中提到: 】
: production DB is OLTP. We do want a OLAP to do the reporting. But because
: time constraints, we will only starts with OLTP dbs copying to offline
: servers insteading of doing all reporting on OLTP dbs right now.
: It's easy to get the backup of OLTP dbs and restore them on the offline
: servers, but it seems not very efficient. If we want to make changes on
: offilne server dbs like tuning index better for reporting, it has to be
: every time we restore them from production dbs. So I'm wondering how we
: do the incremental load like sql data compare does
: We do think about the replication but the architecture is not set up right
: now so we couldn't use it.
※ 来源:·BBS 未名空间站 http://mitbbs.com·[FROM: 69.138.]