Despite the worldwide financial meltdown, all talk on the MS BI blogosphere has been about Project Gemini this week. Even though no-one knows much about it – and certainly no-one knows everything about it either, as we’ve been told there are more announcements to come – the reaction has been pretty clear: the back-end technology looks cool, but the strategy of putting Gemini in Excel on the desktop is an invitation to bypass the data warehouse and create spreadmart hell whatever MS says about Sharepoint-based management of Gemini models. I’ve already linked to just about everyone who’s expressed this opinion apart from Mick Horne, whose three posts here, here and here are well worth reading as they put the case against Gemini in its currently proposed form very clearly. Choice quote: "Where is the single version of the truth in this architecture? I’ve just spent 4 years of my life trying to convince users to stop using Excel as a data store and here are Microsoft positively encouraging it. Hell will freeze over before this capability is used responsibly in most organisations".
So what would I like to see MS actually do with Gemini? From the comments on Marco’s blog, at the moment it sounds like the priority is to get the Gemini storage mode working for local cubes and Excel rather than put everything in place for the full server version of Analysis Services. I would prefer the emphasis to be reversed and have Gemini storage mode ready as soon as possible on the server side. It’s clearly going to provide a massive performance boost when it does arrive (and let’s not forget it’s at least two years away), and with other COP databases and data warehouse appliances improving every day there’s going to be significantly more competitive pressure on Analysis Services and SQL Server-based data warehouse projects in the future. I’d hate to see current MS BI shops start abandoning their Analysis Services implementations because they can get better query performance and scalability elsewhere.
But whatever us bloggers say I’m sure we’ll get Gemini in the form we’ve already been shown. The reason is that whatever the rights and wrongs of it from a BI consultant’s point of view, the people who use Excel will definitely want this and so there will be an overwhelming commercial case for it. This kind of desktop, DIY BI is in a way similar to illegal drugs: there are always some people that want it, a certain number of them are always going to do it even though they know they shouldn’t, so you’ve got two choices – either legalise it and then hope to control it, as with Gemini, or throw all your efforts into outlawing it. With the first option you run the risk of encouraging what you wanted to discourage and ending up with a worse problem, with the second option you run the risk of people resenting your rules so much they end up being widely ignored. What’s the best option?
UPDATE: one last link for you – Nigel Pendse of the OLAP Report gives an unreservedly positive review here: