Sow running queries

Sam 2015-09-18 18:23:17

We have an issue in which sometimes 5 to 10% of the workload ( using store procedure) experience a long execution time ( over a minute compared to less than a second). The issue is resolved if we add/drop an existing index. Our guess is that the "bad" workload must be using a bad execution plan that could possibly be triggered by parameter sniffing. does starting the SQL Server service or rebooting the server clears the cache plan and forces the optimizer to recompile the SP and create a new plan ? We did restart the services/server during the incidents, but the "Bad" workload continued to run over a minute. It was fixed after creating/dropping an index. This is SQL Server 2008 R2 hosted on a VM. Could it be a bug with SQL Server R2 ? VM ? Please advice