Thanks for the response. Yes these files are received on mainframe and they are loaded in to ECM using Mainframe.
The challenge is identifying a pdf start and end from the first file. Just going through google and from my colleagues I found that PDF layouts are having shared resources that means it will have a common fonts, header and footer for the complete document. In that case if i try to split using bytes through my ibm utility I think it will be a corrupted file. I have decided to get back to IBM on this query (We are using IBM ECM tool and IBM utility to load the big binary data in to the tool) and we use HP Dialogue tool for creating pdf files.
Hope the above gives some idea, If any one can suggest some solution I will be very grateful. If I get some anything from IBM, I will update the same over here.
I doubted someone will respond to the queries but it is good to see two replies.