This is more to canvas opinion rather that a question per se.
What is your guys strategy in regards to the netweaver products that support your SAP SoH implementation when it comes to database.
Obviously most JAVA based products don't need the capabilities of an in Memory database and also have very small DB footprints,
e.g. SAP portal, Basic PI, Biller Direct, Mobile Gateway, Business objects, etc.
Do you stay "AnyDB" or put everything on HANA, which given the current limitations of 1 Productive HANA DB per Physical server (be it visualized or not) can be pretty wasteful in terms of compute power?
Another sizing question.
Most white papers assume an average IOPS requirement of 1200. Which is pretty small, when compared to traditional SAP requirements where the rule of thumb for IOPS was SAPs*0.54=IOPS.
And it is not surprising given that you no longer need to read from disk, but when it comes to designing the infrastructure behind it, SAP insists on some pretty beefy storage requirements, i.e. dedicated Storage ports for each HANA node. This means that for larger productive landscapes you end up requiring rather large Storage subsystems with high port counts, at the same time the actual number of spindles is quite low.
I can see the need for storage dedicated to HANA so you don't get mixed workloads interfering with the HANA system, but surely sharing bandwidth between HANA instances should be possible given the very low IOPS requirement.