Mainframe now obsolete for data warehousing
Let Intel inherit the earth?
Reg Reader Workshop Many would have argued in the past that a data warehouse was not a "real" data warehouse unless it existed the in big iron mainframe number crunching environment.
A bit of a generalisation maybe, but when you look at where many of the big banks, insurance companies, retailers and telcos stick their transaction data for analysis even today, it is clear that mainframe class equipment has traditionally been a key part of the overall data warehousing landscape since most people can remember.
Recently though, someone trying to sell me on the merits of using scaled-out Intel based architectures running Windows or Linux made the claim that the mainframe was now pretty much obsolete for things like data warehousing.
I could see his point given the way so called commodity hardware platforms and the database management systems that run on them such as SQL Server, Oracle and MySQL have evolved. The argument is that the cost of these is a fraction of the premium you would pay for the mainframe equivalent, and that things like performance, scalability, manageability, and security are now generally on a par.
Shortly afterwards, however, I happened to be on the receiving end of an IBM mainframe data warehousing pitch, which I have to say was pretty convincing. With the introduction of an offload engine called the Integrated Information Processor (zIIP), which allows certain business intelligence type queries to be routed and executed separately and optimally from the main processor, IBM appeared to be fighting back hard - especially when you add some of the commercial elements into the mix.
In terms of business case, for example, you're unlikely to go out and buy a Series z box for data warehousing purposes if you don't already have one, so the real comparison is the incremental investment of implementing a new warehousing solution on an existing mainframe versus the typical situation of buying, installing, and optimising all of the platform components you need for an new Intel based solution. When you look at it this way and try to ignore all of the fudge factors advocates of one approach or the other try to throw into the mix, the numbers seem pretty close.
Before all you fans of other platforms reach for flamethrowers, yes, of course it's not just about mainframe versus Intel. This raises a question, however. How you go about selecting a platform (or platforms) for a data warehousing project given that requirements range from the big to the small to the distributed/federated, and there are so many options available?
If you have been through this process recently, or just have some thoughts on the matter, we'd love to hear from you in the discussion below. ®