{"id":581,"date":"2008-02-05T08:06:00","date_gmt":"2008-02-05T08:06:00","guid":{"rendered":"\/blogs\/kimberly\/post\/Statistics-query-plans-and-are-you-reading-Conors-blog.aspx"},"modified":"2013-01-11T23:12:08","modified_gmt":"2013-01-12T07:12:08","slug":"statistics-query-plans-and-are-you-reading-conors-blog","status":"publish","type":"post","link":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/statistics-query-plans-and-are-you-reading-conors-blog\/","title":{"rendered":"Statistics, query plans, and are you reading Conor’s blog?"},"content":{"rendered":"

\nI know that Paul and I recommended that you subscribe to Conor's blog<\/a>… but have you? He's posted some great details on Partitioning (Part 1<\/a> and Part 2<\/a>) as well as statistics<\/a> and it always reminds me of how much I can learn from other people's perspectives!\n<\/p>\n

\nAnd, just to dove-tail on some of his statistics comments… I, too, have found that as tables get significantly larger AND have non-standard distributions of more than 200 distinct values (and un-even distribution between those values as well), that the optimizer just cannot possibly do a perfect<\/em> job. The only way an optimizer can be good is when it can "find a good<\/u> plan fast<\/u><\/strong><\/em>" (which I first heard from Nigel Ellis (former Development Manager of the Query Processor team) – back when he delivered a Pacific Northwest SQL Server User Group meeting many moons ago). The most important thing to realize is that it's just not possible to waste time to find the absolutely best plan<\/em>… mathematically analyzing all permutations would be prohibitive – you'd have to take a vacation between query executions (wait, that's not a bad idea… I digress <\/em>:).\n<\/p>\n

\nThe point:\n<\/p>\n

    \n
  1. Make sure that statistics are up-to-date (either through the database option: auto update stats OR by manually updating statistics)<\/li>\n
  2. Consider re-evaluating statistics over large tables (and, when poor performance occurs – look at the estimated rows v. the actual rows – if the estimate\/actual are off by a fact of 10, then it could <\/em>be the statistics). I'd try updating stats first and then if that doesn't work, updating with a fullscan. If neither of those work, I'd also re-evaluate other possible indexes (there are some distributions between tables being joined that just can't show a correct correlation between the values when in multiple indexes… sometimes the best index is a multi-column (ie. composite index)). <\/li>\n
  3. Consider breaking very large tables down into smaller chunks (not just <\/em>table index partitioning but possibly Partitioned Tables AND<\/u> <\/strong>Partitioned Views) as this can give the optimizer additional details about partiticularly interesting data sets. Even in SQL Server 2008, statistics are still table-level (filtered indexes can provide some,<\/em> but not complete,<\/em> relief… I'll give more details in a later post) but I'd often argue that some of the best table designs are not just for a single table. Consider the statistical, locking, and indexing implications for mixed workloads against a single table (and the tremendous amount of blocking that could occur in addition to varying access patterns). And, even while 2008 will offer Partition-level lock escalation<\/a>, well-designed tables may not need it! I know I've mentioned this before but different perspectives on statistics, optimizers and the fact that a good optimizer has to be efficient in-and-of itself, remind me of some of the most basic things that are also the most common problems contributing to poor query performance. <\/li>\n<\/ol>\n

    \nReturning to the basics and optimizing a system from the ground up always leads to better scalability!\n<\/p>\n

    \nEnjoy!
    \nkt\n<\/p>\n

    \n <\/p>\n","protected":false},"excerpt":{"rendered":"

    I know that Paul and I recommended that you subscribe to Conor's blog… but have you? He's posted some great details on Partitioning (Part 1 and Part 2) as well as statistics and it always reminds me of how much I can learn from other people's perspectives! And, just to dove-tail on some of his statistics […]<\/p>\n","protected":false},"author":7,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[25,36,58,66,72],"tags":[],"class_list":["post-581","post","type-post","status-publish","format-standard","hentry","category-design","category-indexes","category-resources","category-sql-server-2008","category-statistics"],"_links":{"self":[{"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/posts\/581","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/comments?post=581"}],"version-history":[{"count":0,"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/posts\/581\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/media?parent=581"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/categories?post=581"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.sqlskills.com\/blogs\/kimberly\/wp-json\/wp\/v2\/tags?post=581"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}