We have 2 fact tables in our dataware house. One for transactions of type V, another for transactions of type M. Each table is partitioned by day and there are 90 days online. There are 145 Million rows of type V, 133 million of type M. They are stuctured in exactly the same way (same partition key, same indexes, etc). The only real difference is that the row size of the table of type V is larger (203 bytes) than the row size of the table of type M (141 bytes). We have a query to look for transactions of type V for customer X. Performance is good. We have an "identical" query to look for transactions of type M for customer X. Performance is horrible. The difference between the two is the query against V is using an INDEX RANGE SCAN. The query against M will only use an INDEX FULL SCAN. What could be the reason?
来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/3637/viewspace-765938/,如需转载,请注明出处,否则将追究法律责任。
转载于:http://blog.itpub.net/3637/viewspace-765938/