Page life expectancy increases dramatically from a minute to another
               Hello. I measure PLE on my server and insert them every minute into a table. Now, when I look into the table I just dont know how to interpret the following data. I dont understand how is that possible. It's an sql server bug? or? Can you help me tointerpret that data please? 
  • Moved by Kalman Toth 11 hours 53 minutes ago Not T-SQL
May 22nd, 2015 10:09am

Hmm.. that's interesting.. I have not seen such a signifcant rise and drop in one minute interval..I would think it is a bug but then PLE is only interpration from SQL based on the workload at the given time, so, my best guess would be sql server activity had zero activity in that duration and hence the value rose quite high..

Again, I am not too sure...

btw, I do not think you want to keep querying PLE every one mintue- may 15.30 mins interval look more realistic..

also, do not depend on PLE itself - you need to keep big picture in mind and consider - total server memory,target memory, memor grants pending, free pages stalls, available memory etc.

Free Windows Admin Tool Kit Click here and download it now
May 22nd, 2015 10:20am

I keep data from every minute and after that I make a daily report with MAX,MIN and AVG. I have another column in my table, in that I insert the max value from an PLE cicle ( when the last value is lower than the one before) I insert that second value into a new column and I make my report from that data.
  • Edited by BogdanSlc 16 hours 38 minutes ago
May 22nd, 2015 10:26am

I keep data from every minute and after that I make a daily report with MAX,MIN and AVG. I have another column in my table, in that I insert the max value from an PLE cicle ( when the last value is lower than the one before) I insert that second value into a new column and I make my report from that data.

Hi Bogdan

I suspect this is a bug in your capture or analysis process. 

PLE is a measure of how long, in seconds, that a data page will remain in memory before being pushed out to make room for new pages. So by definition it should only rise a maximum of 60 seconds for every snapshot you take.

As Stan has said this is something we'd usually only measure every 15/30mins.

I'd suggest observing the perfmon counters and seeing if they match what you're capturing (ultimately) in your table

Try these perfmon objects:

Free Windows Admin Tool Kit Click here and download it now
May 22nd, 2015 2:56pm

Hi,

What is version and edition of SQL Server

select @@Version
Does your SQL Server has NUMA nodes

May 22nd, 2015 4:06pm

BogdanSlc

Can you show us how you measure PLE? 

Free Windows Admin Tool Kit Click here and download it now
May 23rd, 2015 2:28am

This topic is archived. No further replies will be accepted.

Other recent topics Other recent topics