Copy-on-write: Difference between revisions

Revision as of 02:44, 10 November 2023

Latest revision as of 03:40, 13 August 2025

(17 intermediate revisions by 11 users not shown)

Line 1:

{{Short description|Programming technique for efficiently duplicating data}}

{{MoreUse citationsdmy neededdates|date=AugustNovember 20202023}}

'''Copy-on-write''' ('''COW'''), sometimesalso referred to ascalled '''implicit sharing'''<ref>{{cite web |title=Implicit Sharing |url=https://doc.qt.io/qt-5/implicit-sharing.html |website=Qt Project |access-date=10 November 2023 |archive-date=8 February 2024 |archive-url=https://web.archive.org/web/20240208175543/https://doc.qt.io/qt-5/implicit-sharing.html |accessurl-datestatus=4 Augustlive 2016}}</ref> or '''shadowing''',<ref>{{cite journal |last=Rodeh |first=Ohad |title=B-Trees, Shadowing, and Clones |journal=ACM Transactions on Storage |volume=3 |issue=4 |date=1 February 2008 |page=1 |citeseerx=10.1.1.161.6863 |s2cid=207166167 |doi=10.1145/1326542.1326544 |url=http://liw.fi/larch/ohad-btrees-shadowing-clones.pdf |access-date=4 August 2016 |url-status=dead |archive-url=https://web.archive.org/web/20170102212904/http://liw.fi/larch/ohad-btrees-shadowing-clones.pdf |archive-date=2 January 2017 |access-date=10 November 2023 }}</ref> is a [[resourceResource management (computing)|resource-management]] technique used in [[computer programming]] to efficiently implement a "duplicate" or "copy" operation on modifiable [[system resource#General resources|resources]]<ref name="Linux">{{cite book |title=Understanding the Linux Kernel |last1=Bovet |first1=Daniel Pierre |last2=Cesati |first2=Marco |date=1 January 2002-01-01 |title=Understanding the Linux Kernel |url=https://books.google.com/books?id=9yIEji1UheIC&q=%22copy%20on%20write%22&pg=PA295 |publisher=O'Reilly Media |isbn=9780596002138 |page=295 |access-date=10 November 2023 |archive-date=15 September 2024 |archive-url=https://web.archive.org/web/20240915132745/https://books.google.com/books?id=9yIEji1UheIC&q=%22copy%20on%20write%22&pg=PA295#v=snippet&q=%22copy%20on%20write%22&f=false |url-status=live }}</ref> (mostused commonlyin memory[[computer pagesprogramming|programming]] to manage shared data efficiently. Instead of copying data right away when multiple programs use it, storagethe sectorssame data is shared between programs until one tries to modify it. If no changes are made, no private copy is created, saving [[System resource#General resources|resources]].<ref name="Linux" /> A copy is only made when needed, ensuring each program has its own version when modifications occur. This technique is commonly applied to memory, files, and data structures).

When a resource will be duplicated (but not modified), it is not necessary to copy the contents of that resource because, with a little care, every copy can share the same stored contents. (The original counts as one copy because it does not matter which one was the original.) When and if a computer program begins to write into any copy (to change or modify it), the computer must first make a non-shared copy by actually copying the contents, and then write into the new copy. Hence the name of the technique. The actual copy operation is deferred until something needs to write into any copy (which might not ever happen before the copy is discarded). Sharing storage in this way can substantially reduce the consumption of resources by unmodified copies, at the cost of adding some overhead (testing and possibly additional storage) to resource-modifying operations.

==In virtual memory management==

Copy-on-write finds its main use in [[operating system]]s, sharing the [[physical memory]] of computers running multiple [[processProcess (computing)|processes]], in the implementation of the [[forkFork (system call)|fork() system call]]. Typically, the new process does not modify any memory and immediately executes a new process, replacing the address space entirely. It would waste processor time and memory to copy all of the old process's memory during the fork only to immediately discard the copy.<ref>{{Cite Insteadbook the|last=Silberschatz new|first=Abraham process|title=Operating continuesSystem runningConcepts from|last2=Galvin, thePeter sameB. physical|last3=Gagne memory,|first3=Greg in|publisher=Wiley a|year=2018 cheap copy|isbn=978-on-write1119456339 copy|edition=10th of the old process's memory, before discarding it.|pages=120–123}}</ref>

Copy-on-write can be implemented efficiently using the [[page table]] by marking certain pages of [[computerComputer storage|memory]] as read-only and keeping a count of the number of references to the page. When data is written to these pages, the operating-system [[kernelKernel (operating system)|kernel]] intercepts the write attempt and allocates a new physical page, initialized with the copy-on-write data, although the allocation can be skipped if there is only one reference. The kernel then updates the page table with the new (writable) page, decrements the number of references, and performs the write. The new allocation ensures that a change in the memory of one process is not visible in another's.{{Citation needed|date=November 2023}}

The copy-on-write technique can be extended to support efficient [[memory allocation]] by keeping one page of [[physical memory]] filled with zeros. When the memory is allocated, all the pages returned refer to the page of zeros and are all marked copy-on-write. This way, physical memory is not allocated for the process until data is written, allowing processes to reserve more virtual memory than physical memory and use memory sparsely, at the risk of running out of virtual address space. The combined algorithm is similar to [[demand paging]].<ref name="Linux" />

Copy-on-write pages are also used in the [[Linux kernel]]'s [[kernelKernel same-page merging|same-page merging]] feature.<ref>{{citeCite web |last=Abbas |first=Ali |title=The Kernel Samepage Merging Process |website=alouche.net |url=http://alouche.net/blog/2011/07/18/the-kernel-samepage-merging-process/ |access-date=4 August 2016 |url-status=usurped |archive-url=https://web.archive.org/web/20160808174912/http://alouche.net/blog/2011/07/18/the-kernel-samepage-merging-process/ |archive-date=8 August 2016 |access-date=10 November 2023 |website=alouche.net}}</ref>

==In software==

{{Expand section|date=October 2017}}

COW is also used in [[libraryLibrary (computer science)|library]], [[applicationApplication software|application]], and [[systemSystem software|system]] code.

===Examples===

The [[stringString (C++)|string]] class provided by the [[C++ standard library]] was specifically designed to allow copy-on-write implementations in the initial C++98 standard,<ref name="meyers">{{cite book |first=Scott |last=Meyers |author-link=Scott Meyers |date=2012 |title=Effective STL |publisher=Addison-Wesley |pages=64–65 |isbn=9780132979184 |url=https://books.google.com/books?id=U7lTySXdFk0C&pg=PT734}}</ref> but not in the newer C++11 standard:<ref>{{cite web |title=Concurrency Modifications to Basic String |url=https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2534.html |website=Open Standards |access-date=10 November 2023 |archive-date=10 November 2023 |archive-url=https://web.archive.org/web/20231110024434/https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2534.html |accessurl-datestatus=13 Februarylive 2015}}</ref>

<syntaxhighlight lang="cpp">

std::string x("Hello");

Line 30 ⟶ 28:

</syntaxhighlight>

In the [[PHP]] programming language, all types except references are implemented as copy-on-write. For example, strings and arrays are passed by reference, but when modified, they are duplicated if they have non-zero reference counts. This allows them to act as value types without the performance problems of copying on assignment or making them immutable.<ref>{{cite web |last1=Pauli |first1=Julien |last2=Ferrara |first2=Anthony |last3=Popov |first3=Nikita |title=Memory management |date=2013 |url=https://www.phpinternalsbook.com/php5/zvals/memory_management.html#reference-counting-and-copy-on-write |website=PhpInternalsBook.com |access-date=201310 November 2023 |archive-date=10 November 2023 |archive-url=https://web.archive.org/web/20231110024435/https://www.phpinternalsbook.com/php5/zvals/memory_management.html#reference-counting-and-copy-on-write |accessurl-datestatus=4 Augustlive 2016}}</ref>

In the [[Qt (software)|Qt]] framework, many types are copy-on-write ("implicitly shared" in Qt's terms). Qt uses atomic [[compare-and-swap]] operations to increment or decrement the internal reference counter. Since the copies are cheap, Qt types can often be safely used by [[multithreadingMultithreading (computer architecture)|multiple threads]] without the need of [[lockLock (computer science)|locking mechanisms]] such as [[mutualMutual exclusion|mutexes]]. The benefits of COW are thus valid in both single- and multithreaded systems.<ref>{{cite web |title=Threads and Implicitly Shared Classes |website=Qt Project |url=https://doc.qt.io/qt-5/threads-modules.html#threads-and-implicitly-shared-classes |access-date=410 AugustNovember 2023 |archive-date=3 December 2023 |archive-url=https://web.archive.org/web/20231203002914/https://doc.qt.io/QT-5/threads-modules.html#threads-and-implicitly-shared-classes |url-status=live 2016}}</ref>

==In computer storage==

COW may also beis used as the underlying mechanism forin [[snapshotfile (computer storage)|snapshots]], such as those providedsystems bylike [[logical volume managementZFS]], file systems such as [[Btrfs]] and [[ZFS]],<ref>{{cite web |last=Kasampalis |first=Sakis |date=2010 |title=Copy-on-Write Based File Systems Performance Analysis and Implementation |lastpage=Kasampalis19 |firsturl=Sakishttps://sakisk.me/files/copy-on-write-based-file-systems.pdf |access-date=201010 November 2023 |pagearchive-date=195 May 2024 |archive-url=https://web.archive.org/web/20240505220510/https://sakisk.me/files/copy-on-write-based-file-systems.pdf |accessurl-datestatus=11live January 2013}}</ref> and database servers such as [[Microsoft SQL Server#Replication Services|Microsoft SQL ServerReFS]]. Typically, the snapshots store only the modified data, and are stored close to the original[[Bcachefs]], soas theywell areas only a weak form ofin [[incrementallogical backupvolume management]] and cannotdatabase substituteservers forsuch aas [[fullMicrosoft backup]].<ref>{{citeSQL webServer#Replication Services|last=ChienMicrosoft |first=TimSQL |title=Snapshots Are NOT Backups |website=OracleServer]].com |publisher=Oracle |url=https://www.oracle.com/database/technologies/rman-fra-snapshot.html |access-date=4 August 2016}}</ref>

In traditional file systems, modifying a file overwrites the original data blocks in place. In a copy-on-write (COW) file system, the original blocks remain unchanged. When part of a file is modified, only the affected blocks are written to new locations, and metadata is updated to point to them, preserving the original version until it’s no longer needed. This approach enables features like [[Snapshot (computer storage)|snapshots]], which capture the state of a file at a specific time without consuming much additional space. Snapshots typically store only the modified data and are kept close to the original. However, they are considered a weak form of [[incremental backup]] and cannot replace a full backup.<ref>{{cite web |last=Chien |first=Tim |title=Snapshots Are NOT Backups |url=https://www.oracle.com/database/technologies/rman-fra-snapshot.html |website=Oracle.com |publisher=Oracle |access-date=10 November 2023 |archive-date=10 November 2023 |archive-url=https://web.archive.org/web/20231110024434/https://www.oracle.com/database/technologies/rman-fra-snapshot.html |url-status=live }}</ref>

==See also==

Line 43:

* [[Memory management]]

* [[Persistent data structure]]

* [[ReFS]]

* [[Wear leveling]]

==References==

{{Reflist}}

==External links==

* {{cite web |title=A history of copy-on-write memory management |url=https://obvious.services.net/2011/01/history-of-copy-on-write-memory.html |website=A keen grasp of the obvious |access-date=18 November 2024 |date=21 January 2011}}

{{File systems}}