Archives of the TeradataForum

Message Posted: Tue, 04 Nov 2003 @ 12:44:21 GMT

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

Thanks for your help,Chris.

But I'm not sure whether I understand your advice correctly. For example, if one of my universe is composed of one fact table and several dimension table which are

rated_cdr_voice_mfct(fact table), cde_caldate_wtab,cde_call_type_wtab,
cde_roam_type_wtab, cde_trans_type_wtab,
cde_carrire_wtab, cde_srvc_brand_wtab,
cde_gender_wtab,cde_vocation_wtab,
cde_age_level_wtab
BobJ joins are like
 rated_cdr_voice_mfct.bill_month=cde_caldate_wtab.cal_date
 rated_cdr_voice_mfct.call_type=cde_call_type_wtab.call_type
 rated_cdr_voice_mfct.roam_type=cde_roam_type_wtab.roam_type
 rated_cdr_voice_mfct.trans_type=cde_trans_type_wtab.trans_type
 rated_cdr_voice_mfct.other_carrier_id=cde_carrier_wtab.carrier_id
 rated_cdr_voice_mfct.srvc_brand_id=cde_srvc_brand_wtab.srvc_brand_id
 rated_cdr_voice_mfct.vocation_id=cde_vocation_wtab.vocation_id
 rated_cdr_voice_mfct.gender=cde_gender_wtab.gender
 rated_cdr_voice_mfct.age_level=cde_age_level_wtab.age_level

I notice you mention that 'If you find that your BobJ joins are not covering the entire Primary Index, then your Bobj application is doing full table scans and not taking advantage of any part of the Primary Index'.

Does that mean I'd better choose columns bill_month, call_type, roam_type, trans_type, other_carrier_id, srvc_brand_id, vocation_id,gender,age_level combined as the primary index of rated_cdr_voice_mfct.

This is really our choice. The rows of rated_cdr_voice_mfct are 134 million, and a report involved in 5 dimensions of this star schema runs in approx. 15mins. I don't know if I should choose the 5 dimensions as primary index, if it is, what about another report which choose different dimensions?

Thanks,

Zheng Hongxia

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