Data conversion from mainframe to HP Unix

[James George_1]

Hi

We need some insight on migrating data from Mainframe to HP Unix. Right now the data is in a mainframe / EMC env. We are planning to transfer the data to HP Unix systems in a HP Unix / Hitachi env. We are using a tool called cross-os File Exchange provided by Hitachi to convert the data in a unix format so that the HP Unix can read the main frame data . Has any worked on this kind of a project .

Any inputs will be appreciated.

rgds / James

[Yogeeraj_1]

hi james,

as long as the transfer is within the same endian set, you should not encounter any problem.

In any case, you should prepare a test platform and verify that it works properly

good luck!

kind regards
yogeeraj

[Steven Schweda]

It's just possible that the best method for
moving and/or converting the data might
depend on the data. What _are_ "the data"?
How many data are there (MB, GB, TB, ..., YB)?

If you're just moving the bytes from here to
there, and if the Hitachi software can deal
with both ends, then what could go wrong?

And if the volume is not large, or time is
not critical, FTP or something similar might
also do the job.

[A. Clay Stephenson]

This can be trivially easy or almost impossibly difficult, you simply don't provide enough data (or more accurately meta-data) to say with any degree of confidence. There is really no such thing as a "Unix file format" since the format of the data is determined by the application not the OS. When you say "mainframe" that could well imply EBCDIC text encoding so that the text fields would have to be converted to their ASCII equivalents but typically the binary data also associated with a given file might be left intact, byte-swapped, or converted from one encoding scheme to another (e.g. BCD on one end to 64-bit twos-complement on the other) and floating-point conversion can also require translation. The really bad news is that a different data conversion routine might be required for every file.

[Julio Yamawaki · Accepted solution]

Hi,

I use to convert data from IBM mainframes to unix/windows platform.
Normally we export all mainframe data using a utilty like DITTO and put this data to unix platform in binary format.
In unix, using COBOL programs, we convert this EBCDIC data to ASCII data.
It's a simple program and if you need, I can send a sample for you.

Regards,

[James George_1]

Hi Julio

Could you please send me that.

Thanks you .

Rgds / James

[Julio Yamawaki]

Hi,

I attached a source in COBOL format.
User this in this sequence:
1. Export file in mainframe, in EBCDIC format, using a utility like DITTO - this step is used to get rid of index file
2. Send this file in binary format to UNIX
3. Convert this file to ASCII with dd utility (conv=ascii) - after this conversion, field in COMP format will be invalid
4. Use a program like this example to de-convert this file to the original format, but with other fields in ASCII format

[Julio Yamawaki]

This routine de-convert DECIMAL fields.

[Julio Yamawaki]

This routine de-convert COMP-3 fields.

[Julio Yamawaki]

Also, if you have Oracle, you can use this routine to convert EBCDIC to ASCII.

[James George_1]

Thank you very much Julio. I will try these scripts.

rgds / James

[A. Clay Stephenson]

Actually, you don't need COBOL or a database to do the conversions. Perl's Convert::EBCDIC module will do just fine and Perl is very good at reading text files, fixed-lenght records, and even variable length records. As I tried to tell you earlier, the hard part is not the conversion of the field but rather knowing that for example, the first 30 characters are string data, the next 4 are two's-complement integer, the next 12 are BCD, and so on. What you really have to know are the data that describe the data, ie the meta-data.

Here is a link to someone with a similar problem and the Perl solution.

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=236421&admit=-682735245+1164856151846+28353475