We have a website hosted as an Azure App Service. It connects to a Sql Azure P1 database, which is a geo replicated secondary.
We have this same website hosted on a virtual machine with Rackspace, which connects to a sql server database locally installed on the same VM.
The Azure website is accessible at http://www2.inyopools.com The Rackspace website is accessible at http://www.inyopools.com
The performance of the Azure website is significantly worse than the Rackspace one, especially for pages with a lot of database interaction.
For example: http://www2.inyopools.com/pool_part_cat_Above_Ground_Pools.aspx?bustcache=149151216024 Is much slower than http://www.inyopools.com/pool_part_cat_Above_Ground_Pools.aspx?bustcache=149151222081
(Note: these pages are cached every 15 minutes so change the bustcache query string parameter value to get a fresh copy)
Over 12 seconds vs under 1 second difference!
We notice an exponential increase in time based on how many database calls appear to be needed for the page. So, it seems like the bottleneck or issue is database related.
I really need some help figuring out where to go from here in my debugging of the issue. Are read-only, active geo-replicated databases less performant? This is preventing us from migrating these systems to Azure. We’re open to any suggestions!