Read&write binary float var to file

Question

Hello everybody.
Oracle 11G R 2 on Linux.
My problem.
I have some millions of records (270.000.000)  with numeric data that can be store in binary float variables. Making the database work with so big table convert the database in a slow and hard to manage, because copys and so. The advantage is that this data only need be read, not modified or add news record.
So, Im thinking to store in the file system, with UTL_FILE package, one or more binary files and access data through a algorithm to calculate the position of data inside the file. The data can be stored in binary float vars.
My question.
How I can add this binary data in binary format to the stream of write and read from the stream of data when I write to disk and when I read from disk.
Any idea welcome.
Thanks in advanced & regards everybody.

Answer 1

I think that you dont understand the problem.
The problem is easy, I have a table with 270.000.000 records. When you try to do anything with this table the database go slow, for your understand, when you ask a simple and easy query, by example a select count(*) from, it take more than 200 seconds to answer.
ok, but the data is all numeric that can be stored in binary float var and moreover its position in a sequential file can be calculate.
So if I make a sequential file, or more than one, where I can save this data I kick this problem to the database.
And what I´m asking for help is if anybody know a way, using UTL_FILE, to include in the stream to save and read this data from disk (UTL_FILE.put_raw and UTL_FILE.get_raw) in binary float format, nothing more than this I want.
You are correct that I 'dont understand the problem' because you yourself do not know what the 'problem' is.
All you said is 'try to do anything with this table the database go slow' but you don't tell us what that 'anything' is. You give only ONE example.
A COUNT(*) in Oracle could possibly use an index but you don't tell us if you have any indexes. You don't even provide the table DDL so we can see the structure.
Your description gives the impression that all you have is a table with ONE column defined as BINARY FLOAT and that column is defined as NOT NULL. If that is the case you can create an index and Oracle can use it to get the COUNT(*) results.
That ONE query is just about the only thing that might be faster with the data in a file. Since each value is four bytes and they can't be null then all you need to know is the length of the file. Then you can divide by four to find out how many entries there are.
For anything else your 'file' solution will need to read the ENTIRE FILE to do anything other than access values as if the file was a large array of BINARY FLOAT four byte values.
And for that, as I previously suggested, just forget Oracle and write a simple Java program that uses RandomAccessFile. Then for any given 'array' value just multiply by four, set the FilePointer value to that result and read your 4 bytes.