Speaker 2
is off by default, right? Like we don't even get planning time even in the newer versions that do include planning time, which is not that many. It's off by default. So even the I use Cloud SQL at the moment, and they don't even let you turn it on. So it's quite difficult to get planning information.
Speaker 1
Yep. But I'm talking about this fact that single normalized query can be can contain many cases. Like if you show a simple example, when I usually use a select star from some table where column equals one, or equals two. For example, if 99% of values is one, and only 1% is two, in one case, you will get sequential scan. And in other case, you will get index scan. It's two different cases, very different in terms of performance. And P.G.S.S. statements does allow you to distinguish them at all.
Speaker 2
No good point. So yeah, no example, plans. Lack of
Speaker 1
weight event analysis, because it could could be combined with P.G.S.S. statements. For example, if you have normalized query, you understand that like if you could build some pie chart understanding when we spend time, okay, we have total time, mean time, min, max time, we have this. Some of
Speaker 2
the building blocks. Yeah. But like query, query identifier as well now. Right. In
Speaker 1
some cases, we lose time doing some actual work. It can be CPU, it can be in some cases with just weight being blocked by other sessions. In this case, CPU is free. It can do something else. Right. But where is this for? Right. We don't have this for at all. There are many other things. And I know Perkona tried to address it with P.G.S.S.T. monitor. It's quite good and worth looking. But there overhead, I would definitely measure. As we discussed recently, I would definitely measure and understand before going to production. But it's an interesting alternative to P.G.S.S. statements.
