This refers to the different approach that I´ve referred to with some of your other questions. I spent several years working with IBM mainframes, and then several years as a system architect working in Unix environments, where we developed an approach which is radically different from the IBM mainframe approach.

This was essentially a contradictory approach: instead of one large machine, we used separate machines for each of many (carefully defined) purposes - the object-oriented approach, applied at all levels, including the hardware level. This was associated with Unix, networks, client-server architecture, and its further development is associated with Linux, Java, Grid and the "blade" concept. This kind of thinking is difficult if not impossible, to reconcile with the mainframe architecture traditionally preferred by IBM.

Thus putting the two together is difficult to imagine: it is "counter-intuitive". IBM is nevertheless doing this, because the other approach has been very successful and they have to go along with it. What the sentence is saying is that you might think it´s not possible to use both the blade approach and the mainframe approach to server architecture in the same infrastructure, but they are doing it. (This is because they have to, otherwise their mainframes get left out in the cold in many companies.)