I think the reason is it's writing data line as a hash name and the data line can have 300.000 characters.

No, it's not.

At least, if your description of the file is accurate it isn't.

This bit of the code: $Library_Index{<$Library>} = tell(ARGV), reads the IDs and constructs the hash.

And this bit: scalar <$Library> reads and discards the long data lines.

However, Now I think I see the problem with your version of the code.

This bit:until eof(); of the line iterates until the file is read, except that you forgot to put the filehandle $Library in the parens, so the program will never end because it is testing the end-of-file condition of a different file which will never be true.

Change the line to:

$Library_Index{<$Library>} = tell(ARGV), scalar <$Library> until eof($
+Library);
[download]

And see how long it takes.