I don't start by looking at Page Life Expectancy. That metric made sense in the SQL Server 7/2000 days when it was all that we had, but today, in 2018, we can do better.

Start by looking at wait stats - that tells you what SQL Server is waiting on.

I don't care whether PLE is 300 or 3,000 - tell me what you're WAITING on, SQL Server, and then I'll go troubleshoot that metric.

My personal favorite way to check waits is to use the open source sp_BlitzFirst (disclaimer: I wrote it.) By default, it takes a 5-second sample of your server's metrics and gives you a few guesses as to why it's slow right now.

Because you like writing long questions, you'll probably also like these:

sp_BlitzFirst @SinceStartup = 1;

The first result set gives you your waits since startup, and:

sp_Blitz @ExpertMode = 1, @Seconds = 60;

Takes a longer sample, and tells your waits over that time range.

Wait stats can be kind of cryptic, so next to every wait type, I link to the SQLskills wait stats repository for that wait type. You can just copy/paste out the name of your top wait type, go to their site, and learn more about what causes that wait and how to fix it.

If PLE is dropping due to queries reading a lot of data pages from disk, for example, you might see PAGEIOLATCH% wait types. If it's dropping due to queries getting huge memory grants, you might see RESOURCE_SEMAPHORE. If PLE isn't the problem, then you'll see different wait types altogether.

You basically asked, "What should I do when Page Life Expectancy changes?"

My answer: nothing. I don't start by looking at Page Life Expectancy. That metric made sense in the SQL Server 7/2000 days when it was all that we had, but today, in 2018, we can do better.

Start by looking at wait stats - that tells you what SQL Server is waiting on.

I don't care whether PLE is 300 or 3,000 - tell me what you're WAITING on, SQL Server, and then I'll go troubleshoot that metric.

My personal favorite way to check waits is to use the open source sp_BlitzFirst (disclaimer: I wrote it.) By default, it takes a 5-second sample of your server's metrics and gives you a few guesses as to why it's slow right now.

Because you like writing long questions, you'll probably also like these:

sp_BlitzFirst @SinceStartup = 1;

The first result set gives you your waits since startup, and:

sp_Blitz @ExpertMode = 1, @Seconds = 60;

Takes a longer sample, and tells your waits over that time range.

Wait stats can be kind of cryptic, so next to every wait type, I link to the SQLskills wait stats repository for that wait type. You can just copy/paste out the name of your top wait type, go to their site, and learn more about what causes that wait and how to fix it.

If PLE is dropping due to queries reading a lot of data pages from disk, for example, you might see PAGEIOLATCH% wait types. If it's dropping due to queries getting huge memory grants, you might see RESOURCE_SEMAPHORE. If PLE isn't the problem, then you'll see different wait types altogether.