SELECT cast((1.0*cntr_value/128) /
(SELECT 1.0*cntr_value
FROM sys.dm_os_performance_counters
WHERE object_name like ‘%Buffer Manager%’
AND lower(counter_name) = ‘Page life expectancy’) as DECIMAL(8,2))
AS [BufferPoolRate]
FROM sys.dm_os_performance_counters
WHERE
counter_name = ‘Total Server Memory (KB)’

Mike,

Great feedback, thanks. I like the idea of your approach for getting something closer to a real-time diagnostic tool. As you mentioned, the approach above may not work as well for a system that has been running for a long period of time. However, it would still have some value should it reveal that there are a high number of page splits per batch requests. That’s really the goal here, to understand that page splits or batch requests by themselves may not be useful, but putting the two together would yield something better.

I’m now thinking if your approach would be something perfect for monitoring an Azure VM, as you’d want to be alerted if your I/O was heavy as a result of page splits. Now you’ve got me thinking I need to write (yet another) blog post…someday.

Thanks again for the comment!

Seems better to get the counter, WAITFOR DELAY a second, get the counter again, subtract the two values, and show the number of splits that have happened in the last second. This will be independent of the number of batches processed, though similar math could be done against a second pair of samples to correctly retrieve that information.