Archives of the TeradataForum

Message Posted: Mon, 04 Mar 2002 @ 18:07:40 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Yes it is, but I'm not too sure how excited I would get over that fact.

I realize that you're asking a blue-sky question, but I wouldn't recommend trying to collect statistics on the Dictionary tables and I would hate to think that anybody else might think of it as a whiz-bang idea.

I believe that you need to have DROP TABLE on any table to which you want to collect statistics. To get DROP TABLE on a DBC table, you'll need to logon as DBC and do all the work to bypass access rights. The bottom line is that you won't be able to automate the process.

Definitely don't collect statistics on non-hashed tables. I have no idea what would happen, but it wouldn't surprise me if this didn't give you an opportunity to test your disaster recovery process.

I'm pretty certain that NCR doesn't test with statistics on the Dictionary tables and to do something like this without in-depth testing by the vendor just seems to be asking for a whole lot of unpleasant excitement. If NCR isn't actively testing with statistics on the Dictionary tables, then you're putting yourself into a situation where you're going to find all kinds of new/unexpected and probably nasty bugs. And since you're doing something that they haven't tested or recommended or have established procedures for, you're 100% on your own.

Since most of the activity against the dictionary tables are PI operations, with only a few rows being affected, I'm not sure what having statistics would do for you. You may be able to collect statistics, but I don't think that your going to see much benefit - if any. The couple of times that I have heard of statistics being collected, it was to diagnose or address RDBMS problems that were later fixed (and statistics dropped).

So overall, I wouldn't try it without a better understanding of what you're going to gain versus the risk that you're taking. I wouldn't do anything without getting the blessing of NCR first.

	<Prev	Next>		<<First	<Prev	Next>	Last>>