Strategy on backups, master and named instances

  • I am looking for some strategy regarding spreading tables across instances, databases and filegroups for optimal performance of a 'normal' database.

    A normal database contains tables relatively static, fastgrowing tables and 'normal' tables. Examples of static can be domaintables/lookuptables. Fastgrowing is for example a loggtable. Or 'camels', tables growing fast and shrinking heavily occasionally. Normal tables are, just normal tables.

    I can put fastgrowing tables on another instance, thereby making it possible to backup the transactionlog with another frequency than the other tables.

    I can put fastgrowing tables in another database, thereby making it possible to backup and administer it separately from the other tables.

    I can put fastgrowing tables in a filegroup of its own, thereby making it possible to backup and administer it separately from the other tables.

    What are your opinions/experience in this matter? Pros and cons?

    Also I have heard that it is good practice to put master/model in a filegroup/mdf of its own.

    What are your opinions/experience in that matter?

  • I think that the best choise would be to spread the tables in different filegroups with many files inside the filegroups.

    You could group all the normal tables in one filegroup with 3 files. Others tables in more filegroups with more files.

    I don't think that you will gain performance spreading in differents instances.

Viewing 2 posts - 1 through 1 (of 1 total)

You must be logged in to reply to this topic. Login to reply