The first performance impressions are in and it looks like NAV 2013 is the first version on SQL Server that is going to outperform the native database server.
When talking about whether to upgrade to NAV 2013, one of the things that my customers are interested in most is whether it will perform at acceptable levels. I've been wanting to do something elaborate and use the new testability capabilities in NAV 2013 to do a benchmark test for performance specifically. As often happens, plans don't always come together, and by the time I started forming some ideas, other people have gone and done a much better job than I could have. So, what I thought I'd do is compile a number of links to performance related content that I've found so far.
First off, my fellow MVP Jörg Stryk wrote a very technical article about his impressions of NAV 2013 from a SQL Server point of view. He shares his sources, he's done some comparisons between his own T-SQL and the queries that are generated by NAV 2013. Very interesting to read his train of thought, and, if you can keep up with him, you will learn a thing or two about SQL Server. Some of the key things that I picked up on are the modified locking mechanism, the promise of the query object, and the annoying fact that individual users cannot be identified from the SQL Profiler anymore.
Next, my coworker Faithie writes about some changes that were made to how Dimensions are handled in NAV 2013. This one has a huge potential to help improve performance. I work with customers that have millions of ledger entries, with hundreds of millions of ledger entry dimensions, and the prospect of eliminating pretty much all of those is a very compelling one.
Next up, another fellow MVP, Vjekoslav Babić, who lists his top 5 SQL Server improvements. With NAV 2013 being the first version on SQL Server only, we can finally start letting go of certain "SQL improvements" from previous versions, which were really just things to make NAV on SQL behave the same way as the native database server. If I were to list what I take away from this article, I would list all 5, so I'll spare you and just refer you to the article itself.
Where it gets interesting is when mergetool.com's Per Mogensen puts together a couple of videos (here and here) to demonstrate how NAV 2013 stacks up against NAV 2009. These videos then inspired Vjeko to take it a couple of steps further, and put together his own test, and he even shares his test objects. This is turning into a real community effort!
With NAV 2013, it looks at though the potential is definitely there to get performance up to native database server expectations, which was always the argument against using SQL Server. What I am very much excited about is that finally the NAV team can let go of artificially making NAV on SQL Server behave the same way as the native database server and let SQL Server do what it is good at. I personally think that they have only just scratched the surface, and a ton of further improvements are possible. I think the query object will prove one of the most potent weapons against performance problems, and, hopefully, we'll see a tighter integration of SQL Server capabilities from C/SIDE.