I understand SQL Server doesn't release the memory unless the OS needs it. Then, monitoring Available Bytes (free memory) is not the best way to monitor the service. What other variables can give me a real measure about the behaviour of the SQL Server? Maybe Pages/sec or Page Faults/sec? I'm using nagios to monitor the service and sometimes the alerts arise because a big query is executed.
You can get proper memory usage with SQL using the Perfmon counters, in particular the Total Server Memory. With Nagios you could use NSClient++ and nrpe to poll these counters. Here are some SQL Counters for your reference (With silly warning and critical levels):
Example Command Def:
Lastly, you might be interested in Brent Ozar's recommended SQL Perfmon counters: http://www.brentozar.com/archive/2006/12/dba-101-using-perfmon-for-sql-performance-tuning/