Mainframe, MVS and zOS Discussion

In an interview I was aked if we can move the mainframe to open system architecture? For example, if an application on mainframe has got COBOL and DB2 usage, can we really migrate fully to a Oracle based architecture.

Having briefly worked on a migration project. I quickly said, yes, we can make that move. Then we reached to a stage where we started discussing that if it's doable then why people are still not directly developing in oracle? And what are the challanges for mainframe migration?

To this I replied, that meeting scalibility of mainframe is touch. And here we went in to argument, kind of. That if we get millions of tranactions hitting mainframe and oracle system both, they'll behave nealry same. I was told "prove". Which I could not do.

What is/are your opinion on this subject?

What platform is your "oracle system" running on? Few Intel-based systems (whether Unix or Windows or Linux) will be capable of handling millions of transactions per second; the benchmarks that you see touted in articles are usually highly customized and tweaked hardware that is oversized to reach the rates shown in the benchmark. A typical server install will not be able get anywhere near that rate because the hardware won't support it.

A much more reasonable explanation for why people haven't migrated to Oracle is that migration projects are very risky, very expensive, usually take a number of years, and leave the company behind its competitors who are not spending the resources to migrate. A previous employer of mine decided to get rid of its mainframe about 2008. There were 2 main applications running on the mainframe, and the first moved off the mainframe at the end of 2012 (I heard informally the total cost was close to $10 million). The systems staff sent most of 2013 upgrading the hardware and dealing with end user complaints of slow response times and lack of functionality compared to the mainframe system. The other application had one migration project cancelled, and the last I heard was not planned to move off the mainframe until the summer of 2016 -- I haven't heard yet if it really got moved or not.

"Oracle System" is supposed to be on Windows.

"Let's assume that money is available and decision is made to move then should we move and can we meet the scalability of mainframes in Windows oracle system?"... is a question I can think of was asked too. Basically, the discussion moved towards "scalibility" if all things are in place.

A z13 can have up to 141 processors and it ranges from 31 MSU to 13,078 MSU (million service units -- roughly speaking, MIPS is about 8 times these numbers). What a Windows oracle system scale is, I have no idea since I've never had any reason to look at such machines. You really need to find a Windows oracle forum to ask your questions; we focus on mainframes here.

Thanks Robert.

I alos never was part of such migration project from end to end but in some of the companies I have worked with there were such migration programs butas you say, it looks like this not usually a good decision?

it looks like this not usually a good decision?

The specifics will depend upon the company and its mainframe use, of course, but you hear a LOT more about failed / cancelled migration projects than you hear about successful migrations. Managers see the amount being spent on the mainframe and want to get rid of that cost, not realizing that they easily can double or triple the amount for servers to do the same amount of work. Plus mainframe software costs are typically high (it's not unusual to buy a mainframe software product for over $100,000 and maintenance runs 15 to 20% of the purchase price) compared to individual server software costs. Of course, if you need 16 times the base price for the 16 servers you replaced your mainframe with, the server software costs can start adding up as well.