Commodity computing: Difference between revisions

Revision as of 17:49, 26 December 2024

Latest revision as of 03:57, 13 August 2025

(One intermediate revision by one other user not shown)

Line 13:

Such systems are said to be based on standardized computer components, since the standardization process promotes lower costs and less differentiation among vendors' products. Standardization and decreased differentiation lower the switching or exit cost from any given vendor, increasing purchasers' leverage and preventing [[vendor lock-in|lock-in]].

A governing principle of commodity computing is that it is preferable to have more low-performance, low-cost hardware working in parallel (scalar computing) (e.g. [[Advanced Micro Devices|AMD]] x86 [[Complex instruction set computing|CISC]]<ref>{{Cite web|url=http://www.computerworld.com/s/article/9154518/IBM_HP_servers_won_t_stop_x86_onslaught_on_Unix|title = IBM, HP servers won't stop x86 onslaught on Unix|date = 9 February 2010}}</ref>) than to have fewer high-performance, high-cost hardware items<ref>{{Cite web|url=http://research.google.com/pubs/DistributedSystemsandParallelComputing.html|title=Publications – Google Research}}</ref> (e.g. IBM [[POWER7]] or [[Sun Microsystems|Sun]]-[[Oracle Corporation|Oracle's]] [[SPARC]]<ref>[ftp://ftp.software.ibm.com/common/ssi/pm/rg/n/poo03017usen/POO03017USEN.PDF ftp.software.ibm.com]{{dead link|date=AugustMay 20242025|bot=medic}}{{cbignore|bot=medic}}</ref> [[Reduced instruction set computing|RISC]]). At some point, the number of discrete systems in a cluster will be greater than the [[mean time between failures]] (MTBF) for any hardware platform{{Dubious |sentence on fault tolerance/mtbf doesn't make sense|reason=Units don't match for compared items: 1st one is without unit and for the second one unit is time|date=September 2017}}, no matter how reliable, so [[fault tolerance]] must be built into the controlling software.<ref>{{Cite journal|doi = 10.2200/S00193ED1V01Y200905CAC006|title = The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines|year = 2009|last1 = Barroso|first1 = Luiz André|last2 = Hölzle|first2 = Urs|journal = Synthesis Lectures on Computer Architecture|volume = 4|pages = 1–108|doi-access = free}}</ref><ref>{{Cite web |url=http://insidehpc.com/2008/06/02/google-fellow-sheds-some-light-on-infrastructure-robustness-in-face-of-failure |title=Google Fellow sheds some light on infrastructure, robustness in face of failure &#124; insideHPC.com |access-date=2010-03-06 |archive-url=https://web.archive.org/web/20110810085127/http://insidehpc.com/2008/06/02/google-fellow-sheds-some-light-on-infrastructure-robustness-in-face-of-failure/ |archive-date=2011-08-10 |url-status=dead }}</ref>

Purchases should be optimized on cost-per-unit-of-performance, not just on absolute performance-per-CPU at any cost.{{citation needed|date=April 2017}}

Line 58:

*[https://web.archive.org/web/20100209210545/http://labs.google.com/papers/mapreduce-osdi04-slides/index-auto-0021.html Fault tolerance Handled via re-execution]

*[http://hadoop.apache.org/ HADOOP]

*[httphttps://google-services.blogspot.com/2006/07/google-machine.html Google Commodity computing models]

*[https://web.archive.org/web/20160327235023/http://enterprisesystemsmedia.com/article/big-lie-revealed-commodity-servers-not-cheaper-than-mainframe The Big Lie Revealed]