Archives of the TeradataForum

Message Posted: Wed, 13 Aug 2003 @ 20:30:25 GMT

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

Hi Steve and Shelley,

Sounds like you are on the right track. I have just one comment on your recent email.

I'd be a little careful about setting RP0 assigned weight too high, as it could limit the level of diversity you can expect in your lower-weighted RPs. This might be OK if you only plan for a single active group with that RP, or if RP1 was not active at the same time as RP2. As with so much about priority scheduler, it depends.

So let me just illustrate the limitations you could face using the RP weights you outline below (1000, 100, 33).

Using those RP weights, the relative weight that will be attributed to each RP, assuming all 3 are active, will be:

RP0 - 88% (1000 / (1000+100+33) )
RP1 - 8%
RP2 - 2%

So if you wanted to have two AGs active in RP2, you might end up with a relative weight of 1% for each, since the RP only as 2% to divvy up, giving you no differentiation. Or 2% and 0% (with the 0% being interpreted as 1% at run time). But it would be very difficult, with that weight contrast at the RP level to have 3 or 4 active groups (haing different priorities) in RP2. They might all end up with a relative weight of 0%.

An equally conservative approach to selecting RP weights that you might want to consider would be to sum the assigned weights the non- default RPs (133 in your case), then double that number and make that the RP0 assigned weight. The goal is for AG 4 in RP0 to have the highest relative weight, but it doesn't have to be 10 times higher than user allocation groups. Just a bit higher. You'll know if you should reconsider the contrast of weights at the RP level if you start seeing relative weights of 0% for allocation groups showing up on your schmon monitor output.

Thanks, --Carrie

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