Huge matrix distributed unevenly on different processors!

[yafshar]

I have a huge matrix on the order of million times million elements size.
The matrix is being used in a time integration approach and on each processor I have some number of rows of the whole matrix.

For example the whole matrix is 1,000,000*1,000,000 using four processors
0 has 270,000 rows of the whole matrix
1 has 150,000 rows of the whole matrix
2 has 300,000 rows of the whole matrix
3 has 380,000 rows of the whole matrix

Am I able to use SCALAPACK to solve this problem? What approach should I take?
Should I send the information to other processors and make the size even?