How to manage IDENTITY in tables with large number of rows?

[Goran]

Table structure: col1 IDENTITY (seed=1 increment=1) + few other columns (col2...col7) + one text column (col 8)
I have around 50,000,000 rows per day inserted in the table T1. At the end of the day 40,000,000 rows are deleted. I have to keep the records for 12 months and then archive it. Database is 24/7 web serving and there is no down time allowed. IDENTITY column will go out of range (overflow) after less than two years, unless the identity seed is reset to the start value (seed=1, increment=1).
At the end of 12th month data is archived in another table and only last month is kept in the table T1. So table T1 enters new year with data from last month of the previous year. There are few other tables that refer to this table by using there own field with values from T1.IDENTITY column (referential integrity is not enforced). Identity column in T1 is needed as a unique id for some search actions. Performance is an issue therefore bigint data type is used for this identity column rather than decimal.

Another problem I have is how to do table update on one column (1 mil rows to be updated out of 2 mil of rows) with the minimum impact on the users who are querying this table heavily. Not need to mention that it is web app 24/7 no down time.

Thank you in advance.


Goran

[Ray Miao]

Use bigint data type.


------------
Goran at 5/23/2002 6:24:52 PM


Table structure: col1 IDENTITY (seed=1 increment=1) + few other columns (col2...col7) + one text column (col 8)
I have around 50,000,000 rows per day inserted in the table T1. At the end of the day 40,000,000 rows are deleted. I have to keep the records for 12 months and then archive it. Database is 24/7 web serving and there is no down time allowed. IDENTITY column will go out of range (overflow) after less than two years, unless the identity seed is reset to the start value (seed=1, increment=1).
At the end of 12th month data is archived in another table and only last month is kept in the table T1. So table T1 enters new year with data from last month of the previous year. There are few other tables that refer to this table by using there own field with values from T1.IDENTITY column (referential integrity is not enforced). Identity column in T1 is needed as a unique id for some search actions. Performance is an issue therefore bigint data type is used for this identity column rather than decimal.

Another problem I have is how to do table update on one column (1 mil rows to be updated out of 2 mil of rows) with the minimum impact on the users who are querying this table heavily. Not need to mention that it is web app 24/7 no down time.

Thank you in advance.


Goran