You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: 读书笔记/CMU 15-445:645 Database Systems (Fall 2021)/Lecture #11: Query Execution I.md
+1-2Lines changed: 1 addition & 2 deletions
Original file line number
Diff line number
Diff line change
@@ -82,8 +82,7 @@ An access method is how the DBMS accesses the data stored in a table. In general
82
82
83
83
A sequential table scan is almost always the least efficient method by which a DBMS may execute a query. There are a number of optimizations available to help make sequential scans faster:
84
84
85
-
***Prefetching**: Fetch the next few pages in advance so that the DBMS does not have to block on storage
86
-
I/O when accessing each page.
85
+
***Prefetching**: Fetch the next few pages in advance so that the DBMS does not have to block on storage I/O when accessing each page.
87
86
***Buffer Pool Bypass**: The scan operator stores pages that it fetches from disk in its local memory instead of the buffer pool in order to avoid sequential flooding.
88
87
***Parallelization**: Execute the scan using multiple threads/processes in parallel.
89
88
***Zone Map**: ==Pre-compute aggregations== for each tuple attribute in a page. The DBMS can then decide whether it needs to access a page by checking its Zone Map first. The Zone Maps for each page are stored in separate pages and there are typically multiple entries in each Zone Map page. Thus, it is
Copy file name to clipboardExpand all lines: 读书笔记/CMU 15-445:645 Database Systems (Fall 2021)/Lecture #12: Query Execution II.md
+3-1Lines changed: 3 additions & 1 deletion
Original file line number
Diff line number
Diff line change
@@ -158,7 +158,9 @@ To get around this, DBMSs use I/O parallelism to *split installation across mult
158
158
159
159
In multi-disk parallelism, the OS/hardware is configured to store the DBMS’s files across multiple storage devices. This can be done through storage appliances or RAID configuration. All of the storage setup is transparent to the DBMS so workers cannot operate on different devices because the DBMS is unaware of the underlying parallelism.
0 commit comments