Next I’ll reduce the (artificial) page count by half:<\/p>\n

 \r\nUPDATE STATISTICS dbo.[member]\r\nWITH ROWCOUNT = 10000, PAGECOUNT = 74500;\r\n<\/pre>\n
Re-executing the SELECT statement – the Clustered Index Scan operator shows the following estimated costs:<\/p>\n
\n
Estimated CPU Cost = 0.011157 (same as previous measurement of 0.011157)<\/li>\n
Estimated I\/O Cost = 55.1987 (decreased from 110.373)<\/li>\n<\/ul>\n
Inflating the row count while maintaining the page count<\/strong><\/p>\n
What happens if I inflate (unnaturally) the member table’s row count while maintaining the page count?<\/p>\n
 \r\n-- Please don't use this in production - for demonstration purposes only!\r\nUPDATE STATISTICS dbo.[member]\r\nWITH ROWCOUNT = 100000000, PAGECOUNT = 142;\r\n<\/pre>\n
Re-executing the SELECT statement – the Clustered Index Scan operator shows the following estimated costs:<\/p>\n
\n
Estimated CPU Cost = 110<\/li>\n
Estimated I\/O Cost = 0.107569 (no change from the original estimate with 142 data pages and 10,000 rows)<\/li>\n<\/ul>\n
Reducing the row count by half, I see the following estimated costs:<\/p>\n
\n
Estimated CPU Cost = 55.0002<\/li>\n
Estimated I\/O Cost = 0.107569<\/li>\n<\/ul>\n
Summary<\/strong><\/p>\n
So for this post I demonstrated two straight-forward scenarios:<\/p>\n
\n
I\/O Cost of a Clustered Index Scan sensitivity to page counts, but not row counts.<\/li>\n
CPU Cost of a Clustered Index Scan sensitivity to row counts, but not data page counts.<\/li>\n<\/ul>\n
Why care? \u00a0Regarding page counts – think about the estimated I\/O cost impact that high fragmentation may have (same number of rows across many partially filled data pages). \u00a0Regarding row counts – even for narrow rows – consider the \u00a0estimated\u00a0CPU cost impact that high row counts may have.<\/p>\n
 <\/p>\n
 <\/p>\n","protected":false},"excerpt":{"rendered":"
In this post I’ll step through a few scenarios regarding row and page counts – and their associated influence on estimated CPU and I\/O cost. \u00a0I’ll be using the Credit database to demonstrate different scenarios… Starting State I’ll start off by checking the total number of reported data pages and row counts for the member […]<\/p>\n","protected":false},"author":7,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[28],"tags":[],"class_list":["post-815","post","type-post","status-publish","format-standard","hentry","category-performance"],"yoast_head":"\n