1<?xml version="1.0" encoding="UTF-8" standalone="no"?> 2<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> 3<html xmlns="http://www.w3.org/1999/xhtml"> 4 <head> 5 <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /> 6 <title>Locking granularity</title> 7 <link rel="stylesheet" href="gettingStarted.css" type="text/css" /> 8 <meta name="generator" content="DocBook XSL Stylesheets V1.73.2" /> 9 <link rel="start" href="index.html" title="Berkeley DB Programmer's Reference Guide" /> 10 <link rel="up" href="lock.html" title="Chapter��15.�� The Locking Subsystem" /> 11 <link rel="prev" href="lock_deaddbg.html" title="Deadlock debugging" /> 12 <link rel="next" href="lock_notxn.html" title="Locking without transactions" /> 13 </head> 14 <body> 15 <div class="navheader"> 16 <table width="100%" summary="Navigation header"> 17 <tr> 18 <th colspan="3" align="center">Locking granularity</th> 19 </tr> 20 <tr> 21 <td width="20%" align="left"><a accesskey="p" href="lock_deaddbg.html">Prev</a>��</td> 22 <th width="60%" align="center">Chapter��15.�� 23 The Locking Subsystem 24 </th> 25 <td width="20%" align="right">��<a accesskey="n" href="lock_notxn.html">Next</a></td> 26 </tr> 27 </table> 28 <hr /> 29 </div> 30 <div class="sect1" lang="en" xml:lang="en"> 31 <div class="titlepage"> 32 <div> 33 <div> 34 <h2 class="title" style="clear: both"><a id="lock_page"></a>Locking granularity</h2> 35 </div> 36 </div> 37 </div> 38 <p>With the exception of the Queue access method, the Berkeley DB access methods 39do page-level locking. The size of pages in a database may be set when 40the database is created by calling the <a href="../api_reference/C/dbset_pagesize.html" class="olink">DB->set_pagesize()</a> method. If 41not specified by the application, Berkeley DB selects a page size that will 42provide the best I/O performance by setting the page size equal to the 43block size of the underlying file system. Selecting a smaller page size 44can result in increased concurrency for some applications.</p> 45 <p>In the Btree access method, Berkeley DB uses a technique called lock coupling 46to improve concurrency. The traversal of a Btree requires reading a 47page, searching that page to determine which page to search next, and 48then repeating this process on the next page. Once a page has been 49searched, it will never be accessed again for this operation, unless a 50page split is required. To improve concurrency in the tree, once the 51next page to read/search has been determined, that page is locked and 52then the original page lock is released atomically (that is, without 53relinquishing control of the lock manager). When page splits become 54necessary, write locks are reacquired.</p> 55 <p>Because the Recno access method is built upon Btree, it also uses lock 56coupling for read operations. However, because the Recno access method 57must maintain a count of records on its internal pages, it cannot 58lock-couple during write operations. Instead, it retains write locks 59on all internal pages during every update operation. For this reason, 60it is not possible to have high concurrency in the Recno access method 61in the presence of write operations.</p> 62 <p>The Queue access method uses only short-term page locks. That is, a page 63lock is released prior to requesting another page lock. Record locks are 64used for transaction isolation. The provides a high degree of concurrency 65for write operations. A metadata page is used to keep track of the head 66and tail of the queue. This page is never locked during other locking or 67I/O operations.</p> 68 <p>The Hash access method does not have such traversal issues, but it must 69always refer to its metadata while computing a hash function because it 70implements dynamic hashing. This metadata is stored on a special page 71in the hash database. This page must therefore be read-locked on every 72operation. Fortunately, it needs to be write-locked only when new pages 73are allocated to the file, which happens in three cases:</p> 74 <div class="itemizedlist"> 75 <ul type="disc"> 76 <li>a hash bucket becomes full and needs to split</li> 77 <li>a key or data item is too large to fit on a normal page</li> 78 <li>the number of duplicate items for a fixed key becomes so large that they 79are moved to an auxiliary page</li> 80 </ul> 81 </div> 82 <p>In this case, the access method must obtain a write lock on the metadata 83page, thus requiring that all readers be blocked from entering the tree 84until the update completes.</p> 85 <p>Finally, when traversing duplicate data items for a key, the lock on 86the key value also acts as a lock on all duplicates of that key. 87Therefore, two conflicting threads of control cannot access the same 88duplicate set simultaneously.</p> 89 </div> 90 <div class="navfooter"> 91 <hr /> 92 <table width="100%" summary="Navigation footer"> 93 <tr> 94 <td width="40%" align="left"><a accesskey="p" href="lock_deaddbg.html">Prev</a>��</td> 95 <td width="20%" align="center"> 96 <a accesskey="u" href="lock.html">Up</a> 97 </td> 98 <td width="40%" align="right">��<a accesskey="n" href="lock_notxn.html">Next</a></td> 99 </tr> 100 <tr> 101 <td width="40%" align="left" valign="top">Deadlock debugging��</td> 102 <td width="20%" align="center"> 103 <a accesskey="h" href="index.html">Home</a> 104 </td> 105 <td width="40%" align="right" valign="top">��Locking without transactions</td> 106 </tr> 107 </table> 108 </div> 109 </body> 110</html> 111