Introduction Tree-Structured Indexes • Recall: 3 alternatives for data entries k*: CS 186, Fall 2002, Lecture 17 • Data record with key value k R & G Chapter 9 • <k, rid of data record with search key value k> • <k, list of rids of data records with search key k> • Choice is orthogonal to the indexing technique used to locate data entries k*. “If I had eight hours to chop down a • Tree-structured indexing techniques support tree, I'd spend six sharpening my ax.” both range searches and equality searches. • ISAM: static structure; B+ tree: dynamic, Abraham Lincoln adjusts gracefully under inserts and deletes. • ISAM = ???Indexed Sequential Access Method index entry Range Searches P K P K P K P 0 1 1 2 2 m m • ``Find all students with gpa > 3.0’’ ISAM – If data is in sorted file, do binary search to find first such student, then scan to find others. • Index file may still be quite large. But we can – Cost of binary search in a database can be quite apply the idea repeatedly! high. Q: Why??? • Simple idea: Create an `index’ file. Non-leaf Pages k1 k2 kN Index File Leaf Pages Page N Data File Page 1 Page 2 Page 3 Overflow page Primary pages Can do binary search on (smaller) index file! Leaf pages contain data entries. Example ISAM Tree Data Pages ISAM is a STATIC Structure • Index entries:<search key value, page id> • File creation: Leaf (data) pages allocated Index Pages they direct search for data entries in leaves. sequentially, sorted by search key; then • Example where each node can hold 2 entries; index pages allocated, then overflow pgs. Overflow pages • Search: Start at root; use key comparisons to go to leaf. Cost = log F N ; Root F = # entries/pg (i.e., fanout), N = # leaf pgs 40 – no need for `next-leaf-page’ pointers. (Why?) • Insert: Find leaf that data entry belongs to, 20 33 51 63 and put it there. Overflow page if necessary. • Delete: Find and remove from leaf; if empty page, de-allocate. 10* 15* 20* 27* 33* 37* 40* 46* 51* 55* 63* 97* Static tree structure: inserts/deletes affect only leaf pages. 1 Example: Insert 23*, 48*, 41*, 42* ... then Deleting 42*, 51*, 97* Root Root Index 40 Index 40 Pages Pages 20 33 51 63 20 33 51 63 Primary Primary Leaf Leaf 10* 15* 20* 27* 33* 37* 40* 46* 51* 55* 63* 97* 10* 15* 20* 27* 33* 37* 40* 46* 51* 55* 63* 97* Pages Pages Overflow 23* 48* 41* Overflow 23* 48* 41* Pages Pages 42* 42* Note that 51* appears in index levels, but not in leaf! B+ Tree: The Most Widely Used Index ISAM ---- Issues? • Insert/delete at log F N cost; keep tree height- balanced. (F = fanout, N = # leaf pages) • Pros • Minimum 50% occupancy (except for root). Each – ???? node contains d <= m <= 2d entries. “d” is called the order of the tree. • Supports equality and range-searches efficiently. • As in ISAM, all searches go from root to leaves, but • Cons structure is dynamic. – ???? Index Entries (Direct search) Data Entries ("Sequence set") Example B+ Tree B+ Trees in Practice • Search begins at root, and key comparisons • Typical order: 100. Typical fill-factor: 67%. direct it to a leaf (as in ISAM). – average fanout = 133 • Search for 5*, 15*, all data entries >= 24* ... • Typical capacities: 4 Root – Height 4: 133 = 312,900,700 entries – Height 3: 1333 = 2,352,637 entries 13 17 24 30 • Can often hold top levels in buffer pool: – Level 1 = 1 page = 8 Kbytes – Level 2 = 133 pages = 1 Mbyte 2* 3* 5* 7* 14* 16* 19* 20* 22* 24* 27* 29* 33* 34* 38* 39* – Level 3 = 17,689 pages = 133 MBytes Based on the search for 15*, we know it is not in the tree! 2 Inserting a Data Entry into a B+ Tree Example B+ Tree - Inserting 8* • Find correct leaf L. • Put data entry onto L. Root Root –If L has enough space, done! 17 13 17 24 30 –Else, must split L (into L and a new node L2) • Redistribute entries evenly, copy up middle key. 5 13 24 30 • Insert index entry pointing to L2 into parent of L. 2* 3* 5* 7* 14* 16* 19* 20* 22* 24* 27* 29* 33* 34* 38* 39* • This can happen recursively 2* 3* 5*7* 8* 14* 16* 19* 20* 22* 24* 27* 29* 33* 34* 38* 39* – To split index node, redistribute entries evenly, but push up middle key. (Contrast with leaf splits.) Notice that root was split, leading to increase in height. • Splits “grow” tree; root split increases height. In this example, we can avoid split by re-distributing – Tree growth: gets wider or one level taller at top. entries; however, this is usually not done in practice. Example: Data vs. Index Page Split Deleting a Data Entry from a B+ Tree 3* 5* Data 2* 7* 8* Entry to be inserted in parent node. • Observe how Page 5 (Note that 5 iss copied up and • Start at root, find leaf L where entry belongs. minimum Split continues to appear in the leaf.) occupancy is • Remove the entry. guaranteed in 2* 3* 5* 7* 8* – If L is at least half-full, done! both leaf and … –If L has only d-1 entries, index pg splits. •Try to re-distribute, borrowing from sibling (adjacent • Note difference Index 5 13 17 24 30 between copy- Page node with same parent as L). Entry to be inserted in parent node. up and push-up; Split (Note that 17 is pushed up and only 17 appears once in the index. Contrast • If re-distribution fails, merge L and sibling. be sure you this with a leaf split.) understand the • If merge occurred, must delete entry (pointing to L reasons for this. or sibling) from parent of L. 5243013 • Merge could propagate to root, decreasing height. Example Tree (including 8*) Delete 19* and 20* ... ... And Then Deleting 24* • Must merge. 30 Root • Observe `toss’ of Root 17 17 index entry (on right), and `pull down’ of 22* 27* 29* 33* 34* 38* 39* 5 13 24 30 index entry (below). 5 13 27 30 2* 3* 5*7* 8* 14* 16* 19* 20* 22* 24* 27* 29* 33* 34* 38* 39* 2* 3* 5*7* 8* 14* 16* 22* 24* 27* 29* 33* 34* 38* 39* Root 5 13 17 30 • Deleting 19* is easy. • Deleting 20* is done with re-distribution. Notice how middle key is copied up. 2* 3* 5*7* 8* 14* 16* 22* 27* 29* 33* 34* 38* 39* 3 Example of Non-leaf Re-distribution After Re-distribution • Tree is shown below during deletion of 24*. (What • Intuitively, entries are re-distributed by `pushing could be a possible initial tree?) through’ the splitting entry in the parent node. • In contrast to previous example, can re-distribute • It suffices to re-distribute index entry with key 20; entry from left child of root to right child. we’ve re-distributed 17 as well for illustration. Root Root 17 22 5 13 20 22 30 5 13 17 20 30 2* 3* 5*7* 8* 14* 16* 17* 18* 20*21* 22* 27* 29* 33* 34* 38* 39* 2* 3* 5*7* 8* 14* 16* 17* 18* 20* 21* 22* 27* 29* 33* 34* 38* 39* Prefix Key Compression Bulk Loading of a B+ Tree • Important to increase fan-out. (Why?) • If we have a large collection of records, and we • Key values in index entries only `direct traffic’; want to create a B+ tree on some field, doing so can often compress them. by repeatedly inserting records is very slow. – Also leads to minimal leaf utilization --- why? – E.g., If we have adjacent index entries with search key values Dannon Yogurt, David Smith and • Bulk Loading can be done much more efficiently. Devarakonda Murthy, we can abbreviate David Smith • Initialization: Sort all data entries, insert pointer to Dav. (The other keys can be compressed too ...) to first (leaf) page in a new (root) page. • Is this correct? Not quite! What if there is a data entry Root Davey Jones? (Can only compress David Smith to Davi) Sorted pages of data entries; not yet in B+ tree • In general, while compressing, must leave each index entry greater than every key value (in any subtree) to its left. • Insert/delete must be suitably modified. 3* 4* 6* 9* 10* 11* 12* 13* 20* 22* 23* 31* 35* 36* 38* 41* 44* Bulk Loading (Contd.) Summary of Bulk Loading Root 10 20 • Index entries for leaf pages always entered Data entry pages 6 12 23 35 • Option 1: multiple inserts. into right-most index not yet in B+ tree page just above leaf –Slow. level. When this fills – Does not give sequential storage of leaves. up, it splits. (Split 3* 4* 6* 9* 10* 11* 12* 13* 20*22* 23* 31* 35*36* 38*41* 44* may go up right-most • Option 2: Bulk Loading path to the root.) – Has advantages for concurrency control. Root • Much faster than 20 – Fewer I/Os during build. repeated inserts, especially when one 10 35 Data entry pages – Leaves will be stored sequentially (and linked, of considers locking! not yet in B+ tree course). 6 12 23 38 – Can control “fill factor” on pages. 3* 4* 6* 9* 10* 11* 12* 13* 20* 22* 23* 31* 35* 36* 38* 41* 44* 4 A Note on `Order’ Summary • Order (d) concept replaced by physical space • Tree-structured indexes are ideal for range- criterion in practice (`at least half-full’).