SQL Server Performance

Page Life Expectancy

Discussion in 'SQL Server 2005 Performance Tuning for DBAs' started by tfeldman, Feb 25, 2009.

  1. tfeldman New Member

    I have frequently seen 300 sec. as the threshold value for PLE, beyond which you should be concerned about memory pressure. Where does this value of 300 sec come from? Why wouldn't 60 sec or 1 minute be a reasonable length of time for pages to remain in memory?
    Thanks!
  2. FrankKalis Moderator

    Where have you seen this recommendation?
  3. thilina New Member

    Here
    http://sqlcat.com/top10lists/archiv...performance-issues-for-oltp-applications.aspx
    Here
    http://blogs.msdn.com/sqlcat/archive/2006/06/23/Tom-Davidson-SQLCAT-Best-Practices.aspx
    Here
    http://www.microsoft.com/communitie...36-bd71-8bdf1def3342&cat=&lang=&cr=&sloc=&p=1
    Here
    http://www.sqlmag.com/Articles/ArticleID/43117/43117.html?Ad=1
    (This an article named "Page Life Expectancy a Reliable Indicator of SQL Server Memory Pressure" By Brian Moran. I'm not sure without a registration you can view the article, So I pasted some of the content below)
    "Have you ever checked out the page life expectancy counter in Performance Monitor's Buffer Manager object? SQL Server Books Online (BOL) says the page life expectancy value is the "number of seconds a page will stay in the buffer pool without references." So, a buffer that has a 300-second page life expectancy will keep any given page in memory in the buffer pool for 5 minutes before the buffer pool flushes the page to disk—unless a process references the page.
    This counter can be helpful in determining whether you have a memory problem, giving you a reasonably accurate view of whether your server has memory pressure. According to Microsoft, 300 seconds is the minimum target for page life expectancy. If the buffer pool flushes your pages in less than 300 seconds, you probably have a memory problem. Looking at this value is particularly handy when your page life expectancy is significantly higher or lower than 300 seconds. For example, a customer recently asked me, "Do we need more memory?" I monitored the page life expectancy value for the customer's system, and the value didn't fall below 3000 seconds. That's quantifiable proof that more memory wouldn't help performance"
    Here
    http://www.sqlservercentral.com/art...analyzingmemoryrequirementsforsqlserver/2498/

Share This Page