This post also applies to NAV2013, any version since we moved to the new SQL client and dropped classic client.
I am currently in progress of upgrading two of my oldest customers. One with 10 and one with 15 years of data. Large databases, highly customised and tuned for performance to perfection what can be done in those versions.
One of the things you promise these customers is better performance, since NAV2013 and beyond are the fastest versions of Navision ever.
However, after the initial data upgrade and making all the pages more or less usable this is not just the case. Actually the performance is completely terrible. Rubbish.
Ai ai.
The reasoning behind this is relatively simple. Microsoft has made a couple of changes to the way it talks and works with SQL Server that make old performance tuning tips work less and new tips to be required.
First of all we have the “OUTER APPLY”. This is an actual JOIN between flowfields and tables that happens when running pages. Instead of fetching flowfields row by row Navision now creates one query that joins all the flowfields. This can cause the system to slow down if SQL is not optimised for joins.
Learning to optmise SQL for joins is basic knowlegde for DBA’s. Something I learned from sessions organised by SQL Skills. Its all about creating indexes that have the fields from the where clause. Covering indexes also make huge differences.
Second we have the statistics. It used to be best practice to turn off auto create statistics beacuse Navision used cursors. Well, it does no longer do that. We can now safely turn on the default values and not worry about updating them asynchronously.
Lastly there is paralelism on the SQL Box. This used to be set to 1 for NAV databases. Don’t do that anymore. Because of the “OUTER APPLY” we actually bennefit from multiple CPU treads.
Lessons learned? Well this is not my first NAV2013 implementation/upgrade. I learned all this the hard way while emailing back and forth with my former SQL Perform friends. I actually think I had one of the first 100GB+ NAV2013 databases after an early upgrade.
NAV2013 and newer are the fastest versions ever, but performance tuning is an art. It always was and always will be.
Can you screw it up? Yes off course. As a user could Navigate to an Item Ledger Entry in classic client, remove the filter and place a new one and blow up the system, in NAV2013 they can accidentaly sort on an unindexed column. In the last scenario SQL Server will generate and use statistics, but statistics cannot replace indexes.
Enjoy this summer read.
