sql - How would you implement a very wide "table"? -

Suppose you are modeling an entity that has many attributes (2400+), which gives a given physical Database Engine (such as ~ 1000 SQL Server)

a) EAV: Do not know anything about the significance of these data points (people who are often hot / used) domains / Except the candidate key (Boo ... Original Relational Tools window was thrown out.)

B) Go directly across the first table in the primary key and there are 1000 columns, which are right to the limit, the next table is 1000, the foreigner The first table is closed the remaining table is 400, the foreign is also closed.

c) Roof (n / limit) Evenly stripe in the tables is the number of columns in each table, foreign keying for the first table.

And why?

EDIT: This is more of a philosophical / general question, and not linked to any specific limit or engine. Edit

Edit ^ 2: As many have stated, the data was probably not normalized, in general, business barriers make timely research impossible.

My solution: Check further, especially to determine that the table is actually normal ( It is not highly unlikely at 2400 columns).

If not, reconstruction (at that point is likely to be less than 1000 columns per table) until completely normalized.

If it is already completely normal, then install estimated occurrences (as far as possible) for each attribute. Keep the highest attributes on the "home" table for the unit, use 2 or 3 extra tables for less frequently populated features. (Try to create an instance of events, try to make criteria for determining which fields should be played on which tables).

Consider EAV for only very few populated features (preferably, not at all).