Headquarters and Field Sales Sort
2 posters
Page 1 of 1
Headquarters and Field Sales Sort
Hi,
I was reading one of Kimball's articles in the Kimball Group Reader and got somewhat curious about something he says ("7.1 Letting the Users Sleep", Ralph Kimball, DBMS, Dec 1996 and Jan 1997):
"The physical sort order on the disk can be a significant design tool in a data warehouse. In
my data warehouse design classes, I discuss a “headquarters sort” and a “field sales sort.”
The data warehouse designer usually must choose one of these sorts at the expense of the
other. In extreme circumstances, the designer can affect performance by a factor of four
or more, depending on this choice."
Can someone help me understand what is Kimball mentioning when he refers to the “headquarters sort” and “field sales sort.”?
Is this have something to do with partitioning or indexing?
Thanks.
I was reading one of Kimball's articles in the Kimball Group Reader and got somewhat curious about something he says ("7.1 Letting the Users Sleep", Ralph Kimball, DBMS, Dec 1996 and Jan 1997):
"The physical sort order on the disk can be a significant design tool in a data warehouse. In
my data warehouse design classes, I discuss a “headquarters sort” and a “field sales sort.”
The data warehouse designer usually must choose one of these sorts at the expense of the
other. In extreme circumstances, the designer can affect performance by a factor of four
or more, depending on this choice."
Can someone help me understand what is Kimball mentioning when he refers to the “headquarters sort” and “field sales sort.”?
Is this have something to do with partitioning or indexing?
Thanks.
apc- Posts : 8
Join date : 2012-10-11
Re: Headquarters and Field Sales Sort
Considering the age of the article, it probably has to do with clustered index tables. Partitioning does not sort contents of a table.
Given 16 years worth of Moore's Law has passed since then (a 256X or more improvement in performance and capacity and the change to a 64 bit address space), clustered indexes are not as useful as they used to be. Simply providing a bunch of memory on the database server goes a long way to improving performance of sorts and joins.
Given 16 years worth of Moore's Law has passed since then (a 256X or more improvement in performance and capacity and the change to a 64 bit address space), clustered indexes are not as useful as they used to be. Simply providing a bunch of memory on the database server goes a long way to improving performance of sorts and joins.
Similar topics
» Dimensional modelling for Sales and Marketing (opportunity, quotes, sales orders)
» Sales Rep <--> Customer relationship with Sales Fact Table
» Modeling Invoice Level Sales With a Volatile Sales Org
» Trying to design Sales to Promos to Sales Relationship help Please.
» Data model for Sales Order and Sales
» Sales Rep <--> Customer relationship with Sales Fact Table
» Modeling Invoice Level Sales With a Volatile Sales Org
» Trying to design Sales to Promos to Sales Relationship help Please.
» Data model for Sales Order and Sales
Page 1 of 1
Permissions in this forum:
You cannot reply to topics in this forum
|
|