This
level-5 vital article is rated B-class on Wikipedia's
content assessment scale. It is of interest to the following WikiProjects: | |||||||||||
|
Grid computing was a
good article, but it was removed from the list as it no longer met the
good article criteria at the time. There are suggestions below for improving the article. If you can improve it,
please do; it may then be
renominated. Review: September 14, 2006. |
This article is substantially duplicated by a piece in an external publication. Please do not flag this article as a copyright violation of the following source:
|
In the first paragraph, the text says : "as well as a form of network-distributed parallel processing"
This later assertion is quite frivolous... Although it is considered as a form of distributed computing (which is ALWAYS networked btw else it doesn't exist - saying "networked-distributed ..." is a pleonasm), Grid computing has nothing to do with parallelism, there's no notion of process scheduling, topology design or whatsoever from a software designer/engineer stand point, and does not require any parallel programming language syntax/compiler... Compare Beckerley's Parallel C compiler with traditional C compilers and you'll see the difference, it's spectacular. The later compilers can map parallelism over rings, grids, hypercubes, n-cubes, or any other forms of topologies, however these topologies must be taken into account by the software designer/developer via specific language syntax/instructions: it relies on the fact that multiple CPUs are physically part of the same machine, hence with a much more deterministic approach towards very high performances and availability, thus the existence of parallel compilers.
Here, "Distributed parallelism" is an oxymoron: either it's distributed, either it's parallel, but not both. Yes, one could "distribute" an application designed for a parallel computer, but this type of integration (which I've never seen) is completely irrelevant in defining grid or cloud networking. One could end up with grids of parallel computers if need be, but both technologies are very distinct.
Even though they target the same goal which is performance, by coupling the power of multiple CPUs, comparing distributed computing with parallel computing is a TRAP, where such a comparison can confuse or mislead the reader. Distributed computing could be considered as a cheaper alternative for parallelism, but at the same level as considering one single pipelined multi-threaded SIMD or multi-core MIMD CPUs as alternatives to parallel computing: in both cases, performance is the goal, but it's not parallelism (which will always outperform Distributed Computing given the same number of CPUs).
To vulgarize: one could consider Parallel Computing like brain cells in the same head (number of cells, synaptic connections and underlying structure), whereas Distributed Computing (thus including Grid Computing) can be compared to managing the work amongst a number of those "heads" like project-human-resource management (however each "head" has at most 1-2 "brain cells" - or CPUs - in average, usually it doesn't involve parallel computers), for targeting an optimal dead line.
-- HawkFest ( talk) 19:46, 26 June 2009 (UTC)
I would have to disagree, it is a form of parallel computing. While I agree such a distributed system does have fine grain parallelism i.e. the possibility of each instruction been run in parallel it does exhibit something called coarse grain parallelism. This is where parts of the application are sent to different computers in the Grid and executed concurrently i.e. in parallel.
"it's not parallelism (which will always outperform Distributed Computing given the same number of CPUs)." This is a dangerous assumption, large scale computer systems can suffer from scalability issues. It will largely depend upon the overheads of managing such a system. These can vary vastly from system to system. Grids given their largely distributed nature i.e. loose coupling etc. have high degrees of scalability, though I do not argue that HPC (High Performance Computing) systems don't. It must also be considered that algorithms for fine grained parallelism do not always exist and in these cases distributed efforts will work faster. —Preceding unsigned comment added by 92.41.78.191 ( talk) 15:33, 7 March 2010 (UTC)
Is this the same as distributed computing or is it something else? -- Evercat 10:58 May 1, 2003 (UTC)
Do you really think that the Globus Toolkit has something to do with a swiss department store? Maybe you should remove the link from the Grid computing article.
Corrected Globus link issues, though it's a shame that Globus Toolkit automatically redirects to Globus Alliance, should really be a separate article, or at least covered more fully on the Globus Alliance page. Corrected external Globus link from 'The Globus (TM) project' to 'The Globus Alliance, in line with external link in Globus Alliance.
The article adresses only one specific use case of grid computing: collection compute power for big problems. Please see e.g. the OGSA-Use-cases, and you will see that this is only one of many, many use cases. You will find the document GFD.29, 'Open Grid Services Architecture Use Cases' on https://forge.gridforum.org/ . Beside the OGSA Use-cases, there are many more use cases around which are even much more general, but they rise the question what a grid is.
That surely is a candidate for language that makes science and technology articles harder to understand than they need be! call a bilateral trenching tool a spade! -- Tarquin 08:47, 17 Feb 2005 (UTC)
Yes it makes sense.... but I often find myself telling non-geeks about grid computing in the hope they'll try something like SETI@home or a similar project. What should I direct them to for a simple introduction? Certainly not this article as it stands! The opening few paragraphs need to be non-technical. Something that my grandmother could feasibly grasp :) -- Tarquin 15:39, 13 Jun 2005 (UTC)
SETI@home was one of the first grids, however, it was not general purpose. That is, the SETI software was "hard-wired" to do one thing only -- perform pattern recognition on radio telescope data. One of the first general purpose grids and the first commercial grid was created by Parabon Computation. -- Wikiant 21:44, 21 August 2005 (UTC)
SETI@home is cited in a couple places on this page as being the first, but I don't believe that is accurate. About SETI@home and SETI@home#Figures says that it was launched May 1999.
There were other non-profit, scientific public distributed computing efforts prior to then, such as distributed.net, which dates back to April 1997 ( see top paragraph).
Even if distributed.net is not the first—it's quite reasonable that other efforts existed, even if they were not non-profit organizations—it seems reasonable to say that SETI@home was not the first. -- Bovineone 01:37, 14 June 2005.
I would recommend you reading "The Evolution of the Grid" by David De Roure , Mark A. Baker , Nicholas R. Jennings , Nigel R. Shadbolt Grid Computing: Making the Global Infrastructure a Reality if you want to find out about the origins of the Grid. [1]. The earliest notions of the Grid are in project's such as CASA project ,I-WAY and FAFNER and the linking of large super-computing centres together. —Preceding unsigned comment added by 92.41.78.191 ( talk) 14:05, 7 March 2010 (UTC)
Was he not the original inventor? --Ryu
You are totally correct. He was the original inventor. Congratulations. --Dustin Katzin
Why would the CPU owner have a right to a result? It's no different than saying that a calculator should have a right to the result of a problem entered. The value lies not in the solution (which exists whether one finds it or not), but in the asking of the appropriate question.
One of the interesting bits about the Grid is that resources are locally controlled and policy is dictated by that local entity. As we move toward the next generation of protocols (and here I'm specifically thinking of WS-Agreement as a basis for interaction negotiation) a resource may indeed only permit others to use it that agree to share results. If the requester doesn't like that model, then he can continue to look for another resource. -- Rw2 20:51, 16 December 2005 (UTC)
Shouldn't Grid Hosting be mentioned somewhere in this article? PrinzPH ( talk) 17:18, 15 December 2008 (UTC)
How reliable is grid computing? What if one of the computers had faulty hardware (memory, CPU etc), would that cause problems?
Grid computing although reliable largely provides a best effort service. If a computer that is part of the Grid fails or if the owner of that computer takes it out of the Grid the rest of the Grid will carry on going. The work that was scheduled to run on that computer will then be rescheduled elsewhere completely transparently. The major issues with Grids though are that they largely offer best effort services. i.e. they do not guarantee when they will finish their computation. There is a lot of effort going into making them more reliable and adding service level agreements in projects such as AssessGrid. —Preceding unsigned comment added by 92.41.78.191 ( talk) 17:11, 7 March 2010 (UTC)
Globus is certainly widely used, but i'm not sure you can claim its the de-facto middleware solution. Rather, its is a _toolkit_ of which many components are used widely. Overall, given the nature of middleware, describing them as single applications as it were, is misleading. Each is made of many components which can be seperately installed. I removed the claim the Globus was the 'core middleware' for European Grids.
Too bad that was removed. Globus is more than just a toolkit. It is also a suite of services that, without any additional programming, provide a complete grid solution for many classes of deployment. -- Rw2 20:51, 16 December 2005 (UTC)
More Globus information about it would probably be best put in its own article at Globus Toolkit. That article could use some significant additions (and maybe splitting into a separate article), since it currently only discusses the Globus Alliance organization. -- Bovineone 06:27, 17 December 2005 (UTC)
That would be fine, except that Globus Toolkit forwards to Globus Alliance and I don't know how to change that. :-( -- Rw2 23:41, 17 December 2005 (UTC)
I've seen grids running on everything from occassionally-connected laptop 'clusters', through to racks of servers. And, I heard rumours of grid-style applications being developed on configurable CPUs (FPGAs and GPUs, ?) with tremendous performance gains possible. Does anyone know more about the later? -- Richard@lbrc.org 09:50, 13 January 2006 (UTC)
So nurg just did a good copyedit, but he changed all instances of Grid to grid. Personally I use a capitalised version (i work on a Grid project) as i find it distinguished it from a geometric grid. I'd be interested to hear other people's opinions. In the Grid field i find both used depending on the organisation, and don't consider grid inherently more accurate than Grid. ora 12:47, 9 March 2006 (UTC)
Unaware of this note in discussion I went in and changed all references to Grid a few days ago. While working "Grid Computing: The Savvy Manager's Guide" we leaned on a linguist to help us verify the correct spelling. -- Rw2 13:15, 27 April 2006 (UTC)
I feel in reference to this that in Grid computing it should be capitalized. It is possible to have the notion of "The Grid" as in "The Internet". To explain the idea further you can get several internets i.e. inter-networks but the large scale computer network we enjoy today is called "The Internet". Grids should be treated in the same fashion.
So i think this article is due a bit of an overhaul. It has good information but in a pretty strange order as it has grown fairly organically, and 'state of the art 2005' is a bit out of date now we are well into 2006. I'd be happy to have a go but i wanted to see if there were any objections first as i think the article will look pretty difefrent afterwards. I will come up with a revised order first and post it here before i go ahead though. —Preceding unsigned comment added by ora ( talk • contribs)
Oh, please do a complete rewrite. The current state is much of a mess. Smoe
OK, so I'll start on a revision on a subpage at Talk:Grid computing/Draft Revision. It will take me a while, and I'll post again here when I have reached a reasonable point. Others please do contribute but discuss it here as well. I won't implement any of it without discussion here first though. ora 08:10, 23 May 2006 (UTC)
That's a lot of external links... Surely some of them should be removed. See WP:EL for the policy. I would try to remove some, but I'm not too familiar with Grid computing. Gflores Talk 05:12, 19 March 2006 (UTC)
Though this article page says clusters should not be confused with grids. Sun Grid documentation no. 817-6117 does not seem to agree.
As per Sun, Grid is a collection of computers, that are capable of performing a task in a collaboration, appearing to the user as a single entity. There are three classes of grids: Cluster grid, Campus grid, Global grid. Though topologically they are same, the geographical proximity between the members of the grid differentiate their classes. In the Cluster grid the member computers are located in the same rack (or a room) and are connected by a high speed LAN, usually a gigabit LAN. The Campus grid's computers are scattered within a building. And, Global grid, as the name suggests, is distributed across the planet, connected by Internet.
The canonical definition is from Foster and SGE/N1 don't match it. -- Rw2 13:13, 27 April 2006 (UTC)
Sun (as with a number of vendors) confuse Grid computing with cluster computing. In reality, cluster computing is really a degenerate case of Grid computing where the complex issues (e.g., distributed ownership, local autonomy, heterogeneity) have been trivially resolved by defining them to be fixed values. For comparison, the higher-end Grids (alas, often still at the "research platform" stage) use clusters as basic components.
-- Donal Fellows 3 July 2006
Sorry about the use of the text from the page; didn't realize that went against policy. I've re-added the text I wrote (and left out the text from the page as well as the banner, of course.) FlyByPC 19:37, 2 Sep 2003 (UTC)
Should we add info on this 'best discovery yet? [2]/ [3]? -- Piotr Konieczny aka Prokonsul Piotrus 20:23, 2 Sep 2004 (UTC)
Just did, see SHGb02 14a and expand. Whether or not this is little green men talking to us, or a computer bug, it is likely to be newsworthy. It will probably be at least as significant as the Wow! signal historically, even if it doesn't pan out. pstudier 21:51, 2 Sep 2004 (UTC)
If you want to join a team for SETI@home, have a look at the Wikipedia team! -- 80.229.152.246 16:04, 29 Mar 2005 (UTC)
Or if your interested in the World Community Grid, join this team Students for a Cure Caleb rosenberg 18:34, 2 April 2006 (UTC)
Should there be a history section? Who was involved in setting it up? (Is David Anderson notable?) When was it conceived? How long did it take to get set up?
The current text indicates that SETI@home is unique in that it uses coherent integration, but all modern microwave and some optical SETI searches, use both coherent and non-coherent integration.
Coherent here really just means the use of discrete Fourier transforms, rather than simply averaging the power. The link on coherent integration is also misleading as the connection between optical coherence and coherent integration is rather tenuous, although optical coherence is a prerequisite for signals to be detected by coherent integration.
In SETI@home, the gaussian search is effectively non-coherent integration and the pulse search certainly uses non-coherent integration.
What's really unique is the number of chirp rates tried, and, in later versions, the search for repeating pulses.
David Woolley 13:11, 11 October 2005 (UTC)
It is clear that the main threats to the project are funding (though there was a debate about this issue in 2002 [4]) and the appearence of other alternative projects (BOINC - though I consider it as a solution instead of a threat). However, I don't agree about the following:
Good arguments. I agree. --Cheers Svest 20:42, 15 October 2005 (UTC) Wiki me up™
The entire "threats to the project" section sounds slightly short of NPOV, as it subtly takes the view of a project member. "What threatens us?" In particular, calling other grid computing projects a threat is at least questionable terminology. -- Mr. Billion 06:53, 14 March 2006 (UTC)
Be afraid, be very afraid!
http://www.f-secure.com/weblog/archives/archive-112005.html#00000724 http://home.fnal.gov/~carrigan/SETI/SETI_Hacker.htm —Preceding unsigned comment added by 195.70.48.242 ( talk • contribs)
I propose to change the referencing style from inline URLs, plus full citation in References, to using Harvard style references inline (with the same full citations). The advantage of this is that it is easier to see which facts come from which sources, which helps in maintaining the sources and means that it is easier for a reader to judge the reliability of statements against their perception of the reliabilty of particular sources.
Note that the article is generally under sourced at the moment (a common problem on Wikipedia).
Up to yesterday, I believe that I was the only person to have contributed inline references, so I could have made the change unilaterally. A couple of references have now been added inline in the direct URL style, but have not yet been properly cited in the References section. See, for example, WP:V and WP:CITE for why full citations are desirable, although it is also worth noting that Berkeley has a real problem with link rot and renaming URLs. -- David Woolley 09:11, 18 December 2005 (UTC)
On February 11, JarlaxleArtemis added a section about "s-23 wiki",
I removed the section. There are thousands of SETI@home teams; this team does not appear to be any more notable than any other team. Note that s-23 wiki used to have its own article, but it was deleted. dbenbenn | talk 10:32, 15 February 2006 (UTC)
The article states that "Grid purists point out that Seti@home is really a distributed computing application as it does not make use of almost any Grid concepts." (Note that SETI is capitalised in the SETI@home article, whereas this article is inconsistent in its usage.)
About the only attempts to distinguish the concepts of grid computing and distributed computing that I have found in this article or in distributed computing are two separate and inconspicuous sentences in this one. Firstly: "Grid computing's focus on the ability to support computation across administrative domains sets it apart from ... traditional distributed computing." Secondly: "One characteristic that currently distinguishes grid computing from distributed computing is the abstraction of a 'distributed resource' into a grid resource." (Currently? Has this changed? Is it expected to?) It is not immediately apparent to me, even after reading this article and the SETI@home article how the project fails to meet these criteria, as these unnamed "purists" claim.
As a general comment, there appears to be a great deal of confusion with respect to the relationship between grid and distributed computing (and, to a lesser extent, clustered computing). It would appear that a definition by delimitation and relation is called for in the introductory definitions in this article and in distributed computing.
LX 16:07, 8 April 2006 (UTC)
Having said that "SETI@home is not a grid project", the article then goes on to say, In a Grid, only the code required for retrieving work and returning results persists on the nodes. Code required to perform the distributed work is sent to the nodes separately. In this way, the nodes of a Grid can be easily reprogrammed.
This is an absolutely perfect description of (my understanding of) the BOINC framework, upon which SETI@home is built!
Consequently, the preceding assertion, "SETI@home's screensaver contains both code to process radio telescope data and code to handle retrieving work and returning results. The two bodies of code are intertwined into a single program.", is just plain wrong.
Indeed, the arrant nonsense of this assertion can clearly be seen in the fact that SETI@home can be run on Linux/UNIX without any graphics capability at all (I run the clients like this myself). Not alone is the infrastructure/management/network client (BOINC) separate from the project client (SETI@home, Predictor@home, climateprediction.net, ...), but the screensaver, if any, is a separate component again.
-- EmmetCaulfield 07:54, 23 April 2006 (UTC)
I have this vague feeling that no one is using grid computing to get any actual science done. If I were to play the devil's advocate, I would say that grid computing just does not make sense in the current computer world, and it may never make sense. It will always be more advantageous to tightly integrate a supercomputer at a single location, than to attempt high levels of computation via widely distributed computers. The overhead and difficulty of maintaining the broad distribution of computing resources will always work in favor of the large computing resource located at a single site. I feel like the article could have a "Criticisms" section mentioning this issue. Westwind273 23:46, 12 January 2007 (UTC)
I completely agree with the initial statement. I also have the less than vague impression that grid computing is not being used for real science, except, of course, paper and thesis production on the subject itself.
As a matter of fact, I would even say that grid computing can be seen as a subtle form of intellectual fraud.
I cannot figure out how can it be a good idea to pay such a tremendous price in security, reliability (sensible data is spread out through possibly unreliable nodes, my computer can be used by loosely authenticated users that I do not know quite well), complexity, to share resources that are cheap and getting cheaper: CPU cycles, memory, storage.
Just to mention one of them: disk space. Terabytes disks are about to be available. If your application needs hundreds of terabytes, or millions of terabytes, will it help to spread it throughout the world? Ok, the networks provide huge amounts of bandwidth, but unfortunately no fiber can make the light travel faster, so geographic delays is something one have to live with, doesn’t matter how many bandwidth the network provides.
SETI@Home is frequently cited as a case of success. In my opinion SETI@Home, or FightAids@HOME , are extremely well succeeded marketing campaigns’, but could hardly be described as scientific achievements. Any way, if they can, these are very particular cases, that cannot be alleged as rule.
If Wikipedia cannot host a discussion on this, the article should point to articles with criticism to the concept of grid computing.
--
Skandor 23:30, 21 January 2007 (UTC)
As far as I can see the following added nothing to the article and, I think, constituted spam. I have removed it to here as it may be possible to turn this list into a sensible paragraph. Andreww 17:27, 15 May 2006 (UTC)
Key vendors in Grid computing, in alphabetical order:
Vendors of related technology (e.g. schedulers and cluster file systems):
While I support removing the commercial links, or maybe moving them to a sub-page, the last revision that wiped out the seti@hoem section, two of the definitions, and the LCG>EGEE section was too much. Maybe consider writing them but don't just blank them without explanation. SETI@home may or may not be a grid by some definitions but it is worth a mention, the definitions were useful on some level (i might have cut down Buyya for self promotion but not removed his work), and the removal of discussion of two major working production grids was silly. If you dislike the delivery, try rewriting the sections, don't just remove info without explanation. ora 13:02, 18 May 2006 (UTC)
Second this response. The basic idea of cleaning up and spam reduction is sound. I'd even agree that each of the targeted sections could use some work. But, unlike the eyeOS sections previously redacted, the targeted sections do also contain some legit commentary worth saving with modification. -- Rw2 16:01, 23 May 2006 (UTC)
Legit comments do not require the publicizing of company names and company/organizations in the general purpose definition section. Putting such attributes in the lists -- away from the definitions is enough. The text must be removed when the naming of projects and companies is used to publicize that company, or a project that points to some company/organization deliberately. That is a type of a SPAM. A general purpose defintion must avoid referncing company names and companies' project names in order to offer them publicity. It makes no difference if SETI@home was first (although it clearly was NOT -- its simply a client/server architecture in a WAN environment -- aka: distributed computing). Why insert ANY company name or company project names into the general purpose definition section? Putting the deserving companies in the "external links" and other related lists (below the definition) is quite sufficient. -- IlonDalon 22:16, 24 May 2006 (UTC)
Removing the spam ORIGIN paragraphs and replacing them with the following general verbiage works for me: Some companies and organizations claim to be the first ones out of the starting gate. There has never been a clear front runner in the grid computing arena. Grid computing evolved inside several companies and organizations (in parallel) simply out of the need for a more effective computing model; one that solves for those specific companies/organizations a particular set of problems and utilizes the "idle cycle stealing" concept in an region wider than a single computing element. -- IlonDalon 11:58, 25 May 2006 (UTC)
There certainly has been a front runner. The folks at the Globus project brought a bunch of technology ideas together (cycle scavenging, cluster management, virtual organizations, data optimization, storage management, open standards, monitoring and half a dozen others) under one banner, coined the term Grid and remain the de facto standard for doing Grid computing. -- Rw2 21:05, 25 May 2006 (UTC)
I know nothing of the space, I am the confused as to the relevance of SLA's Surely Grid Computing can cache and hache making the use of SLA's irrelevant?
I agree. This article disappears up its own I/O port sometimes and needs clarity. I also agree that the "CPU-cycles sharing" argument is overstated in the modern context. User:safeer.always@gmail.com 11:07, 18 Feb 2005 (UTC)
The links in the section Common features are a bit confusing: "Computational Grids" is just a forward to the top page, the "Data Grids" link leads to an empty page (although there is a page named Data Grid), and the Equipment Grids link is dead, too. As a sidenote, I would be very interested in the source of this taxonomy: Are there any papers I can reference?
-- 89.57.183.10 08:35, 1 July 2006 (UTC)
I just looked over the Globus Alliance page and discovered that it has been rewritten and includes now projects that are actually not part of the Globus Alliance. It may also be worthwhile to point out globdev as a way for the community to contribute.
Gregor von Laszewski, Argonne National Laboratory, http://www.cogkit.org
Much of this article is dedicated to lauding how amazing this grid thing is, and there's a whole section advertising various organizations, that's just ridiculous, there is no way this is a Good Article, delete the ads and change the language to neutral as per WP:NPOV. Homestarmy 13:53, 14 September 2006 (UTC)
I was reading through the article and thought that
This technology was mostly abandoned in the 1980s as the administrative and security issues involved in having machines you did not control do your computation were (and are still by some) seen as insurmountable.
was badly worded and could be replaced with
This technology was mostly abandoned in the 1980s, mainly because of the administrative and security issues that arise from distributing data to foreign machines. Data procured in this way could be seen as being potentially erronous. (data consitency cannot be verified from an uncertified source).
and possibly (sorry im just editing as i read, and i may remove this after ive read the article) could include (I've just made this up, but if somebody verify it) something about the theoretical expodential reliability of data as relative to the number of sources... (i guess thats the old 'good vs evil' debate)
Im no expert, if somebody can ratify/expand then commit it.
Also, 'The grid has more potential than the general public believes' stands to be inacurate as the public at large don't know what the grid is, and the ones that do probably heard about it because of its potential.
This article fails to mention Metacomputing, which is (in its modern incarnation) basically the same thing as Grid. The biggest initial competitor to Globus, Legion, isn't mentioned at all. FYI, the way I remember it, NASA's Information Power Grid pre-dated Globus' use of the term Grid. Greg 06:10, 30 October 2006 (UTC) Superscript text
This article is good. Can someone please offer their services to make the Utility computing good as well, even if that means merging with this one? -- GreatTurtle 02:55, 28 May 2007 (UTC)
I forgot to note in my edit summary that I merged CPU scavenging when I did the partial rewrite. -- Beland 02:51, 4 June 2007 (UTC)
By way of cleaning up the Definitions sections, I think these are appropriate as references for the intro (which presents a summary of common definitions), but too verbose to present in detail to readers. -- Beland 05:03, 4 June 2007 (UTC)
The article should be consistent about whether it is "Grid" or "grid". I think it should probably use "Grid" throughout. Also I'm not sure if it is strictly true to speak of Grid computing as being a specialization of distributed computing; a large part of Grid computing is the idea of viewing computing resources as a service - the user should not care how "distributed" the resources are.
There is obviously a large overlap with distributed computing, I'm just not convinced Grid computing can accurately be described as only a subset of distributed computing.
129.215.63.115 08:52, 22 June 2007 (UTC)
This article is out of date. It represents grid computing as it was circa 2002 or so, when grids were still used predominantly for academic and high-performance computing (HPC) applications such as digital animation and special effects, oil and gas exploration, etc.
In the last 5 years, however, grids have moved into the mainstream of computing. Companies like Oracle, IBM and BEA have introduced grid-enabled versions of major mainstream products, including Oracle 10g (relational database) and IBM WebSphere (J2EE middleware). Numerous startups, including DataSynapse, Appistry, Digipede and ActiveGrid have developed new application middleware for grids. Others, like ExaGrid and Amazon, have built grid -based architectures for scaling storage.
Perhaps more important is the fact that utility computing is rapidly moving mainstream, and all serious projects in this area are based in one way or another on grid architectures. Amazon Web Services has been on the market for a year with their S3/EC2 offering. Microsoft recently released the first details of their upcoming Cloud OS. Ebay has announced "Project eBox". Finally, 3Tera has been shipping for a year a grid operating system called AppLogic, which runs and scales mainstream web applications on grids.
All of this is great news, since it shows that grids are finally moving mainstream. Indeed, the opinion that grids are the next major computing architecture, with utility computing being it's killer app, is rapidly gaining momentum in the industry. Naturally, the fact that dozens of teams at different companies are working in parallel to advance the technology also means that there is a great deal of confusion on terminology and concepts, as every team introduces their own by necessity.
Nevertheless, one thing is clear: grid computing is out of the academia, and in the mainstream of computing. We need a comprehensive rewrite of this article to reflect today's state of the technology, architectures, applications and challenges.
As a professional working in this area (I am one of the founders of 3Tera, mentioned above), I have the obvious motivation to invest significant effort in this work. I am also a former physicist who have spent a few years in academia, so I know the difference between original research, review articles and marketing :-). This is a significant undertaking, so I will need help with editing, checking sources, etc., and, of course, with keeping me honest :-)
I would like to hear thoughts from the community here before I embark on this refresh. Maverick61 21:32, 5 August 2007 (UTC)
I propose changing the first bullet point in the article to:
"a collection of independent computers and local clusters combined into a unified computing infrastructure" Jwozniak 20:08, 8 October 2007 (UTC)
I recently sorted the "See also" lists according to blue links, red links, and external links. Donal Fellows pointed out to me that the sorting before was (nearly) alphabetic, which is arguably better. I was confused partly because the alphabetical order had deteriorated somewhat, but I was actually motivated by the recent addition of a commercial link. I was bothered that that might be spam, but I could not easily what the other links were about, especially since the wiki links were indistinguishable from the external links. In the end I removed the three dot.com links, but I also question whether it is useful to have so many red links, or indeed any at all under "See also". I still maintain that a cleanup of the lists is in order, but I certainly don't insist that my way is best. Other opinions? -- Art Carlson 15:24, 8 November 2007 (UTC)
I'm going to contribute by reformatting tables using Help:Sorting for Table feel free to revert Seanwong ( talk) —Preceding undated comment added 01:14, 18 October 2018 (UTC)
Rereading this page, I see that most of the text in it seems to be defining a Grid as doing things like cycle scavenging and stuff like that. That's not what a Grid is about! More traditional supercomputers and Beowulf clusters also fit into the general Grid model, and the interesting use-cases involve coördination of these different resources to solve a problem. For example, using a cycle-scavenging system to check a large set of molecules for whether they look likely to bind to some target protein, and then using HPC to examine in detail the cases that look at least not impossible after the initial scan. (Doing such things often requires resources spread across more than one organization — requiring non-trivial security — and very careful control over QoS so that the coördination works at all...) To my mind, cycle scavenging is just how you build a particular class of (cheap, low interconnect capability, variable availability) cluster. But then I think my view of this is pretty much what Foster and Kesselman's original vision involved. (Grid computing is probably the same as Cloud Computing, and is closely related to SOA.) Donal Fellows ( talk) 23:34, 1 February 2008 (UTC)
CORBA is important for distributed computing in general, but, AFAIK, no Grid middleware is using CORBA at the moment. I thus removed it from the list of Grid APIs. IMHO, Corba is lakking the scalability (mostly in terms of administration/coordination) to be useful for Grid systems. —Preceding unsigned comment added by Andremerzky ( talk • contribs) 10:39, 24 September 2008 (UTC)
ABC@home used to be fast, http://web.archive.org/web/20120520163950/http://boincstats.com/en/stats/41/project/detail should it be added with "as of May 2012"? The numbers I found (might find even higher for this project or another) than the just updated BIONIC and Folding@home I just updated. This would be kind of misleading since the numbers are way lower now (and have been for a while?). Also It's not clear to me if this (or comments like these on talk pages) should go at the top (seems to be almost in that order here). Comp.arch ( talk) 11:58, 23 April 2013 (UTC)
Hello fellow Wikipedians,
I have just added archive links to one external link on
Grid computing. Please take a moment to review
my edit. If necessary, add {{
cbignore}}
after the link to keep me from modifying it. Alternatively, you can add {{
nobots|deny=InternetArchiveBot}}
to keep me off the page altogether. I made the following changes:
When you have finished reviewing my changes, please set the checked parameter below to true to let others know.
This message was posted before February 2018.
After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than
regular verification using the archive tool instructions below. Editors
have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the
RfC before doing mass systematic removals. This message is updated dynamically through the template {{
source check}}
(last update: 5 June 2024).
Cheers. — cyberbot II Talk to my owner:Online 08:06, 28 August 2015 (UTC)
Hello fellow Wikipedians,
I have just modified one external link on Grid computing. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{
Sourcecheck}}
).
This message was posted before February 2018.
After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than
regular verification using the archive tool instructions below. Editors
have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the
RfC before doing mass systematic removals. This message is updated dynamically through the template {{
source check}}
(last update: 5 June 2024).
Cheers.— cyberbot II Talk to my owner:Online 06:15, 30 April 2016 (UTC)
Hello fellow Wikipedians,
I have just modified 6 external links on Grid computing. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.
This message was posted before February 2018.
After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than
regular verification using the archive tool instructions below. Editors
have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the
RfC before doing mass systematic removals. This message is updated dynamically through the template {{
source check}}
(last update: 5 June 2024).
Cheers.— InternetArchiveBot ( Report bug) 05:55, 24 October 2017 (UTC)
An IP has alleged a potential problem. I removed their comment and replaced it with a maintenance tag. Ifnord ( talk) 21:36, 3 June 2020 (UTC)
Heterogeneous computing and grid computing appear to be functionally indistinguishable. Ethanpet113 ( talk) 23:13, 28 August 2020 (UTC)
I am opposed the this merger. Heterogeneous computing is concerned with the micros-scale, within the bounds of the computing node. Grid Computing is on the macro-scale and describes the challenges of federating services across data centers. — Preceding unsigned comment added by 217.140.110.7 ( talk) 19:57, 30 September 2020 (UTC)
I'm also opposed, for the same reason: heterogeneous computing is (simplistically) about stuffing GPGPU cards into servers, whereas Grid is about access to a variety of remote, shared resources. It doesn't help that this article presently fails to mention "virtual organisations" or "single sign-on" which are basically what distinguishes Grid [1] from clouds, clusters and other distributed computing frameworks. -- Louis Knee ( talk) 17:39, 4 November 2020 (UTC)
Absolutely not. Heterogenous computing and Grid computing are different concepts which should not be confused. While a grid system can be heterogenous (with multiple ISAs), it does not have to be. Neither does a heterogenous computing system need to involve multiple simultaneous networked or even interconnected nodes. Heterogeneity can also be expressed over time ( checkpointing, with resumption on a single node later). Johan Hanson ( talk) 16:50, 15 December 2020 (UTC)
References
This
level-5 vital article is rated B-class on Wikipedia's
content assessment scale. It is of interest to the following WikiProjects: | |||||||||||
|
Grid computing was a
good article, but it was removed from the list as it no longer met the
good article criteria at the time. There are suggestions below for improving the article. If you can improve it,
please do; it may then be
renominated. Review: September 14, 2006. |
This article is substantially duplicated by a piece in an external publication. Please do not flag this article as a copyright violation of the following source:
|
In the first paragraph, the text says : "as well as a form of network-distributed parallel processing"
This later assertion is quite frivolous... Although it is considered as a form of distributed computing (which is ALWAYS networked btw else it doesn't exist - saying "networked-distributed ..." is a pleonasm), Grid computing has nothing to do with parallelism, there's no notion of process scheduling, topology design or whatsoever from a software designer/engineer stand point, and does not require any parallel programming language syntax/compiler... Compare Beckerley's Parallel C compiler with traditional C compilers and you'll see the difference, it's spectacular. The later compilers can map parallelism over rings, grids, hypercubes, n-cubes, or any other forms of topologies, however these topologies must be taken into account by the software designer/developer via specific language syntax/instructions: it relies on the fact that multiple CPUs are physically part of the same machine, hence with a much more deterministic approach towards very high performances and availability, thus the existence of parallel compilers.
Here, "Distributed parallelism" is an oxymoron: either it's distributed, either it's parallel, but not both. Yes, one could "distribute" an application designed for a parallel computer, but this type of integration (which I've never seen) is completely irrelevant in defining grid or cloud networking. One could end up with grids of parallel computers if need be, but both technologies are very distinct.
Even though they target the same goal which is performance, by coupling the power of multiple CPUs, comparing distributed computing with parallel computing is a TRAP, where such a comparison can confuse or mislead the reader. Distributed computing could be considered as a cheaper alternative for parallelism, but at the same level as considering one single pipelined multi-threaded SIMD or multi-core MIMD CPUs as alternatives to parallel computing: in both cases, performance is the goal, but it's not parallelism (which will always outperform Distributed Computing given the same number of CPUs).
To vulgarize: one could consider Parallel Computing like brain cells in the same head (number of cells, synaptic connections and underlying structure), whereas Distributed Computing (thus including Grid Computing) can be compared to managing the work amongst a number of those "heads" like project-human-resource management (however each "head" has at most 1-2 "brain cells" - or CPUs - in average, usually it doesn't involve parallel computers), for targeting an optimal dead line.
-- HawkFest ( talk) 19:46, 26 June 2009 (UTC)
I would have to disagree, it is a form of parallel computing. While I agree such a distributed system does have fine grain parallelism i.e. the possibility of each instruction been run in parallel it does exhibit something called coarse grain parallelism. This is where parts of the application are sent to different computers in the Grid and executed concurrently i.e. in parallel.
"it's not parallelism (which will always outperform Distributed Computing given the same number of CPUs)." This is a dangerous assumption, large scale computer systems can suffer from scalability issues. It will largely depend upon the overheads of managing such a system. These can vary vastly from system to system. Grids given their largely distributed nature i.e. loose coupling etc. have high degrees of scalability, though I do not argue that HPC (High Performance Computing) systems don't. It must also be considered that algorithms for fine grained parallelism do not always exist and in these cases distributed efforts will work faster. —Preceding unsigned comment added by 92.41.78.191 ( talk) 15:33, 7 March 2010 (UTC)
Is this the same as distributed computing or is it something else? -- Evercat 10:58 May 1, 2003 (UTC)
Do you really think that the Globus Toolkit has something to do with a swiss department store? Maybe you should remove the link from the Grid computing article.
Corrected Globus link issues, though it's a shame that Globus Toolkit automatically redirects to Globus Alliance, should really be a separate article, or at least covered more fully on the Globus Alliance page. Corrected external Globus link from 'The Globus (TM) project' to 'The Globus Alliance, in line with external link in Globus Alliance.
The article adresses only one specific use case of grid computing: collection compute power for big problems. Please see e.g. the OGSA-Use-cases, and you will see that this is only one of many, many use cases. You will find the document GFD.29, 'Open Grid Services Architecture Use Cases' on https://forge.gridforum.org/ . Beside the OGSA Use-cases, there are many more use cases around which are even much more general, but they rise the question what a grid is.
That surely is a candidate for language that makes science and technology articles harder to understand than they need be! call a bilateral trenching tool a spade! -- Tarquin 08:47, 17 Feb 2005 (UTC)
Yes it makes sense.... but I often find myself telling non-geeks about grid computing in the hope they'll try something like SETI@home or a similar project. What should I direct them to for a simple introduction? Certainly not this article as it stands! The opening few paragraphs need to be non-technical. Something that my grandmother could feasibly grasp :) -- Tarquin 15:39, 13 Jun 2005 (UTC)
SETI@home was one of the first grids, however, it was not general purpose. That is, the SETI software was "hard-wired" to do one thing only -- perform pattern recognition on radio telescope data. One of the first general purpose grids and the first commercial grid was created by Parabon Computation. -- Wikiant 21:44, 21 August 2005 (UTC)
SETI@home is cited in a couple places on this page as being the first, but I don't believe that is accurate. About SETI@home and SETI@home#Figures says that it was launched May 1999.
There were other non-profit, scientific public distributed computing efforts prior to then, such as distributed.net, which dates back to April 1997 ( see top paragraph).
Even if distributed.net is not the first—it's quite reasonable that other efforts existed, even if they were not non-profit organizations—it seems reasonable to say that SETI@home was not the first. -- Bovineone 01:37, 14 June 2005.
I would recommend you reading "The Evolution of the Grid" by David De Roure , Mark A. Baker , Nicholas R. Jennings , Nigel R. Shadbolt Grid Computing: Making the Global Infrastructure a Reality if you want to find out about the origins of the Grid. [1]. The earliest notions of the Grid are in project's such as CASA project ,I-WAY and FAFNER and the linking of large super-computing centres together. —Preceding unsigned comment added by 92.41.78.191 ( talk) 14:05, 7 March 2010 (UTC)
Was he not the original inventor? --Ryu
You are totally correct. He was the original inventor. Congratulations. --Dustin Katzin
Why would the CPU owner have a right to a result? It's no different than saying that a calculator should have a right to the result of a problem entered. The value lies not in the solution (which exists whether one finds it or not), but in the asking of the appropriate question.
One of the interesting bits about the Grid is that resources are locally controlled and policy is dictated by that local entity. As we move toward the next generation of protocols (and here I'm specifically thinking of WS-Agreement as a basis for interaction negotiation) a resource may indeed only permit others to use it that agree to share results. If the requester doesn't like that model, then he can continue to look for another resource. -- Rw2 20:51, 16 December 2005 (UTC)
Shouldn't Grid Hosting be mentioned somewhere in this article? PrinzPH ( talk) 17:18, 15 December 2008 (UTC)
How reliable is grid computing? What if one of the computers had faulty hardware (memory, CPU etc), would that cause problems?
Grid computing although reliable largely provides a best effort service. If a computer that is part of the Grid fails or if the owner of that computer takes it out of the Grid the rest of the Grid will carry on going. The work that was scheduled to run on that computer will then be rescheduled elsewhere completely transparently. The major issues with Grids though are that they largely offer best effort services. i.e. they do not guarantee when they will finish their computation. There is a lot of effort going into making them more reliable and adding service level agreements in projects such as AssessGrid. —Preceding unsigned comment added by 92.41.78.191 ( talk) 17:11, 7 March 2010 (UTC)
Globus is certainly widely used, but i'm not sure you can claim its the de-facto middleware solution. Rather, its is a _toolkit_ of which many components are used widely. Overall, given the nature of middleware, describing them as single applications as it were, is misleading. Each is made of many components which can be seperately installed. I removed the claim the Globus was the 'core middleware' for European Grids.
Too bad that was removed. Globus is more than just a toolkit. It is also a suite of services that, without any additional programming, provide a complete grid solution for many classes of deployment. -- Rw2 20:51, 16 December 2005 (UTC)
More Globus information about it would probably be best put in its own article at Globus Toolkit. That article could use some significant additions (and maybe splitting into a separate article), since it currently only discusses the Globus Alliance organization. -- Bovineone 06:27, 17 December 2005 (UTC)
That would be fine, except that Globus Toolkit forwards to Globus Alliance and I don't know how to change that. :-( -- Rw2 23:41, 17 December 2005 (UTC)
I've seen grids running on everything from occassionally-connected laptop 'clusters', through to racks of servers. And, I heard rumours of grid-style applications being developed on configurable CPUs (FPGAs and GPUs, ?) with tremendous performance gains possible. Does anyone know more about the later? -- Richard@lbrc.org 09:50, 13 January 2006 (UTC)
So nurg just did a good copyedit, but he changed all instances of Grid to grid. Personally I use a capitalised version (i work on a Grid project) as i find it distinguished it from a geometric grid. I'd be interested to hear other people's opinions. In the Grid field i find both used depending on the organisation, and don't consider grid inherently more accurate than Grid. ora 12:47, 9 March 2006 (UTC)
Unaware of this note in discussion I went in and changed all references to Grid a few days ago. While working "Grid Computing: The Savvy Manager's Guide" we leaned on a linguist to help us verify the correct spelling. -- Rw2 13:15, 27 April 2006 (UTC)
I feel in reference to this that in Grid computing it should be capitalized. It is possible to have the notion of "The Grid" as in "The Internet". To explain the idea further you can get several internets i.e. inter-networks but the large scale computer network we enjoy today is called "The Internet". Grids should be treated in the same fashion.
So i think this article is due a bit of an overhaul. It has good information but in a pretty strange order as it has grown fairly organically, and 'state of the art 2005' is a bit out of date now we are well into 2006. I'd be happy to have a go but i wanted to see if there were any objections first as i think the article will look pretty difefrent afterwards. I will come up with a revised order first and post it here before i go ahead though. —Preceding unsigned comment added by ora ( talk • contribs)
Oh, please do a complete rewrite. The current state is much of a mess. Smoe
OK, so I'll start on a revision on a subpage at Talk:Grid computing/Draft Revision. It will take me a while, and I'll post again here when I have reached a reasonable point. Others please do contribute but discuss it here as well. I won't implement any of it without discussion here first though. ora 08:10, 23 May 2006 (UTC)
That's a lot of external links... Surely some of them should be removed. See WP:EL for the policy. I would try to remove some, but I'm not too familiar with Grid computing. Gflores Talk 05:12, 19 March 2006 (UTC)
Though this article page says clusters should not be confused with grids. Sun Grid documentation no. 817-6117 does not seem to agree.
As per Sun, Grid is a collection of computers, that are capable of performing a task in a collaboration, appearing to the user as a single entity. There are three classes of grids: Cluster grid, Campus grid, Global grid. Though topologically they are same, the geographical proximity between the members of the grid differentiate their classes. In the Cluster grid the member computers are located in the same rack (or a room) and are connected by a high speed LAN, usually a gigabit LAN. The Campus grid's computers are scattered within a building. And, Global grid, as the name suggests, is distributed across the planet, connected by Internet.
The canonical definition is from Foster and SGE/N1 don't match it. -- Rw2 13:13, 27 April 2006 (UTC)
Sun (as with a number of vendors) confuse Grid computing with cluster computing. In reality, cluster computing is really a degenerate case of Grid computing where the complex issues (e.g., distributed ownership, local autonomy, heterogeneity) have been trivially resolved by defining them to be fixed values. For comparison, the higher-end Grids (alas, often still at the "research platform" stage) use clusters as basic components.
-- Donal Fellows 3 July 2006
Sorry about the use of the text from the page; didn't realize that went against policy. I've re-added the text I wrote (and left out the text from the page as well as the banner, of course.) FlyByPC 19:37, 2 Sep 2003 (UTC)
Should we add info on this 'best discovery yet? [2]/ [3]? -- Piotr Konieczny aka Prokonsul Piotrus 20:23, 2 Sep 2004 (UTC)
Just did, see SHGb02 14a and expand. Whether or not this is little green men talking to us, or a computer bug, it is likely to be newsworthy. It will probably be at least as significant as the Wow! signal historically, even if it doesn't pan out. pstudier 21:51, 2 Sep 2004 (UTC)
If you want to join a team for SETI@home, have a look at the Wikipedia team! -- 80.229.152.246 16:04, 29 Mar 2005 (UTC)
Or if your interested in the World Community Grid, join this team Students for a Cure Caleb rosenberg 18:34, 2 April 2006 (UTC)
Should there be a history section? Who was involved in setting it up? (Is David Anderson notable?) When was it conceived? How long did it take to get set up?
The current text indicates that SETI@home is unique in that it uses coherent integration, but all modern microwave and some optical SETI searches, use both coherent and non-coherent integration.
Coherent here really just means the use of discrete Fourier transforms, rather than simply averaging the power. The link on coherent integration is also misleading as the connection between optical coherence and coherent integration is rather tenuous, although optical coherence is a prerequisite for signals to be detected by coherent integration.
In SETI@home, the gaussian search is effectively non-coherent integration and the pulse search certainly uses non-coherent integration.
What's really unique is the number of chirp rates tried, and, in later versions, the search for repeating pulses.
David Woolley 13:11, 11 October 2005 (UTC)
It is clear that the main threats to the project are funding (though there was a debate about this issue in 2002 [4]) and the appearence of other alternative projects (BOINC - though I consider it as a solution instead of a threat). However, I don't agree about the following:
Good arguments. I agree. --Cheers Svest 20:42, 15 October 2005 (UTC) Wiki me up™
The entire "threats to the project" section sounds slightly short of NPOV, as it subtly takes the view of a project member. "What threatens us?" In particular, calling other grid computing projects a threat is at least questionable terminology. -- Mr. Billion 06:53, 14 March 2006 (UTC)
Be afraid, be very afraid!
http://www.f-secure.com/weblog/archives/archive-112005.html#00000724 http://home.fnal.gov/~carrigan/SETI/SETI_Hacker.htm —Preceding unsigned comment added by 195.70.48.242 ( talk • contribs)
I propose to change the referencing style from inline URLs, plus full citation in References, to using Harvard style references inline (with the same full citations). The advantage of this is that it is easier to see which facts come from which sources, which helps in maintaining the sources and means that it is easier for a reader to judge the reliability of statements against their perception of the reliabilty of particular sources.
Note that the article is generally under sourced at the moment (a common problem on Wikipedia).
Up to yesterday, I believe that I was the only person to have contributed inline references, so I could have made the change unilaterally. A couple of references have now been added inline in the direct URL style, but have not yet been properly cited in the References section. See, for example, WP:V and WP:CITE for why full citations are desirable, although it is also worth noting that Berkeley has a real problem with link rot and renaming URLs. -- David Woolley 09:11, 18 December 2005 (UTC)
On February 11, JarlaxleArtemis added a section about "s-23 wiki",
I removed the section. There are thousands of SETI@home teams; this team does not appear to be any more notable than any other team. Note that s-23 wiki used to have its own article, but it was deleted. dbenbenn | talk 10:32, 15 February 2006 (UTC)
The article states that "Grid purists point out that Seti@home is really a distributed computing application as it does not make use of almost any Grid concepts." (Note that SETI is capitalised in the SETI@home article, whereas this article is inconsistent in its usage.)
About the only attempts to distinguish the concepts of grid computing and distributed computing that I have found in this article or in distributed computing are two separate and inconspicuous sentences in this one. Firstly: "Grid computing's focus on the ability to support computation across administrative domains sets it apart from ... traditional distributed computing." Secondly: "One characteristic that currently distinguishes grid computing from distributed computing is the abstraction of a 'distributed resource' into a grid resource." (Currently? Has this changed? Is it expected to?) It is not immediately apparent to me, even after reading this article and the SETI@home article how the project fails to meet these criteria, as these unnamed "purists" claim.
As a general comment, there appears to be a great deal of confusion with respect to the relationship between grid and distributed computing (and, to a lesser extent, clustered computing). It would appear that a definition by delimitation and relation is called for in the introductory definitions in this article and in distributed computing.
LX 16:07, 8 April 2006 (UTC)
Having said that "SETI@home is not a grid project", the article then goes on to say, In a Grid, only the code required for retrieving work and returning results persists on the nodes. Code required to perform the distributed work is sent to the nodes separately. In this way, the nodes of a Grid can be easily reprogrammed.
This is an absolutely perfect description of (my understanding of) the BOINC framework, upon which SETI@home is built!
Consequently, the preceding assertion, "SETI@home's screensaver contains both code to process radio telescope data and code to handle retrieving work and returning results. The two bodies of code are intertwined into a single program.", is just plain wrong.
Indeed, the arrant nonsense of this assertion can clearly be seen in the fact that SETI@home can be run on Linux/UNIX without any graphics capability at all (I run the clients like this myself). Not alone is the infrastructure/management/network client (BOINC) separate from the project client (SETI@home, Predictor@home, climateprediction.net, ...), but the screensaver, if any, is a separate component again.
-- EmmetCaulfield 07:54, 23 April 2006 (UTC)
I have this vague feeling that no one is using grid computing to get any actual science done. If I were to play the devil's advocate, I would say that grid computing just does not make sense in the current computer world, and it may never make sense. It will always be more advantageous to tightly integrate a supercomputer at a single location, than to attempt high levels of computation via widely distributed computers. The overhead and difficulty of maintaining the broad distribution of computing resources will always work in favor of the large computing resource located at a single site. I feel like the article could have a "Criticisms" section mentioning this issue. Westwind273 23:46, 12 January 2007 (UTC)
I completely agree with the initial statement. I also have the less than vague impression that grid computing is not being used for real science, except, of course, paper and thesis production on the subject itself.
As a matter of fact, I would even say that grid computing can be seen as a subtle form of intellectual fraud.
I cannot figure out how can it be a good idea to pay such a tremendous price in security, reliability (sensible data is spread out through possibly unreliable nodes, my computer can be used by loosely authenticated users that I do not know quite well), complexity, to share resources that are cheap and getting cheaper: CPU cycles, memory, storage.
Just to mention one of them: disk space. Terabytes disks are about to be available. If your application needs hundreds of terabytes, or millions of terabytes, will it help to spread it throughout the world? Ok, the networks provide huge amounts of bandwidth, but unfortunately no fiber can make the light travel faster, so geographic delays is something one have to live with, doesn’t matter how many bandwidth the network provides.
SETI@Home is frequently cited as a case of success. In my opinion SETI@Home, or FightAids@HOME , are extremely well succeeded marketing campaigns’, but could hardly be described as scientific achievements. Any way, if they can, these are very particular cases, that cannot be alleged as rule.
If Wikipedia cannot host a discussion on this, the article should point to articles with criticism to the concept of grid computing.
--
Skandor 23:30, 21 January 2007 (UTC)
As far as I can see the following added nothing to the article and, I think, constituted spam. I have removed it to here as it may be possible to turn this list into a sensible paragraph. Andreww 17:27, 15 May 2006 (UTC)
Key vendors in Grid computing, in alphabetical order:
Vendors of related technology (e.g. schedulers and cluster file systems):
While I support removing the commercial links, or maybe moving them to a sub-page, the last revision that wiped out the seti@hoem section, two of the definitions, and the LCG>EGEE section was too much. Maybe consider writing them but don't just blank them without explanation. SETI@home may or may not be a grid by some definitions but it is worth a mention, the definitions were useful on some level (i might have cut down Buyya for self promotion but not removed his work), and the removal of discussion of two major working production grids was silly. If you dislike the delivery, try rewriting the sections, don't just remove info without explanation. ora 13:02, 18 May 2006 (UTC)
Second this response. The basic idea of cleaning up and spam reduction is sound. I'd even agree that each of the targeted sections could use some work. But, unlike the eyeOS sections previously redacted, the targeted sections do also contain some legit commentary worth saving with modification. -- Rw2 16:01, 23 May 2006 (UTC)
Legit comments do not require the publicizing of company names and company/organizations in the general purpose definition section. Putting such attributes in the lists -- away from the definitions is enough. The text must be removed when the naming of projects and companies is used to publicize that company, or a project that points to some company/organization deliberately. That is a type of a SPAM. A general purpose defintion must avoid referncing company names and companies' project names in order to offer them publicity. It makes no difference if SETI@home was first (although it clearly was NOT -- its simply a client/server architecture in a WAN environment -- aka: distributed computing). Why insert ANY company name or company project names into the general purpose definition section? Putting the deserving companies in the "external links" and other related lists (below the definition) is quite sufficient. -- IlonDalon 22:16, 24 May 2006 (UTC)
Removing the spam ORIGIN paragraphs and replacing them with the following general verbiage works for me: Some companies and organizations claim to be the first ones out of the starting gate. There has never been a clear front runner in the grid computing arena. Grid computing evolved inside several companies and organizations (in parallel) simply out of the need for a more effective computing model; one that solves for those specific companies/organizations a particular set of problems and utilizes the "idle cycle stealing" concept in an region wider than a single computing element. -- IlonDalon 11:58, 25 May 2006 (UTC)
There certainly has been a front runner. The folks at the Globus project brought a bunch of technology ideas together (cycle scavenging, cluster management, virtual organizations, data optimization, storage management, open standards, monitoring and half a dozen others) under one banner, coined the term Grid and remain the de facto standard for doing Grid computing. -- Rw2 21:05, 25 May 2006 (UTC)
I know nothing of the space, I am the confused as to the relevance of SLA's Surely Grid Computing can cache and hache making the use of SLA's irrelevant?
I agree. This article disappears up its own I/O port sometimes and needs clarity. I also agree that the "CPU-cycles sharing" argument is overstated in the modern context. User:safeer.always@gmail.com 11:07, 18 Feb 2005 (UTC)
The links in the section Common features are a bit confusing: "Computational Grids" is just a forward to the top page, the "Data Grids" link leads to an empty page (although there is a page named Data Grid), and the Equipment Grids link is dead, too. As a sidenote, I would be very interested in the source of this taxonomy: Are there any papers I can reference?
-- 89.57.183.10 08:35, 1 July 2006 (UTC)
I just looked over the Globus Alliance page and discovered that it has been rewritten and includes now projects that are actually not part of the Globus Alliance. It may also be worthwhile to point out globdev as a way for the community to contribute.
Gregor von Laszewski, Argonne National Laboratory, http://www.cogkit.org
Much of this article is dedicated to lauding how amazing this grid thing is, and there's a whole section advertising various organizations, that's just ridiculous, there is no way this is a Good Article, delete the ads and change the language to neutral as per WP:NPOV. Homestarmy 13:53, 14 September 2006 (UTC)
I was reading through the article and thought that
This technology was mostly abandoned in the 1980s as the administrative and security issues involved in having machines you did not control do your computation were (and are still by some) seen as insurmountable.
was badly worded and could be replaced with
This technology was mostly abandoned in the 1980s, mainly because of the administrative and security issues that arise from distributing data to foreign machines. Data procured in this way could be seen as being potentially erronous. (data consitency cannot be verified from an uncertified source).
and possibly (sorry im just editing as i read, and i may remove this after ive read the article) could include (I've just made this up, but if somebody verify it) something about the theoretical expodential reliability of data as relative to the number of sources... (i guess thats the old 'good vs evil' debate)
Im no expert, if somebody can ratify/expand then commit it.
Also, 'The grid has more potential than the general public believes' stands to be inacurate as the public at large don't know what the grid is, and the ones that do probably heard about it because of its potential.
This article fails to mention Metacomputing, which is (in its modern incarnation) basically the same thing as Grid. The biggest initial competitor to Globus, Legion, isn't mentioned at all. FYI, the way I remember it, NASA's Information Power Grid pre-dated Globus' use of the term Grid. Greg 06:10, 30 October 2006 (UTC) Superscript text
This article is good. Can someone please offer their services to make the Utility computing good as well, even if that means merging with this one? -- GreatTurtle 02:55, 28 May 2007 (UTC)
I forgot to note in my edit summary that I merged CPU scavenging when I did the partial rewrite. -- Beland 02:51, 4 June 2007 (UTC)
By way of cleaning up the Definitions sections, I think these are appropriate as references for the intro (which presents a summary of common definitions), but too verbose to present in detail to readers. -- Beland 05:03, 4 June 2007 (UTC)
The article should be consistent about whether it is "Grid" or "grid". I think it should probably use "Grid" throughout. Also I'm not sure if it is strictly true to speak of Grid computing as being a specialization of distributed computing; a large part of Grid computing is the idea of viewing computing resources as a service - the user should not care how "distributed" the resources are.
There is obviously a large overlap with distributed computing, I'm just not convinced Grid computing can accurately be described as only a subset of distributed computing.
129.215.63.115 08:52, 22 June 2007 (UTC)
This article is out of date. It represents grid computing as it was circa 2002 or so, when grids were still used predominantly for academic and high-performance computing (HPC) applications such as digital animation and special effects, oil and gas exploration, etc.
In the last 5 years, however, grids have moved into the mainstream of computing. Companies like Oracle, IBM and BEA have introduced grid-enabled versions of major mainstream products, including Oracle 10g (relational database) and IBM WebSphere (J2EE middleware). Numerous startups, including DataSynapse, Appistry, Digipede and ActiveGrid have developed new application middleware for grids. Others, like ExaGrid and Amazon, have built grid -based architectures for scaling storage.
Perhaps more important is the fact that utility computing is rapidly moving mainstream, and all serious projects in this area are based in one way or another on grid architectures. Amazon Web Services has been on the market for a year with their S3/EC2 offering. Microsoft recently released the first details of their upcoming Cloud OS. Ebay has announced "Project eBox". Finally, 3Tera has been shipping for a year a grid operating system called AppLogic, which runs and scales mainstream web applications on grids.
All of this is great news, since it shows that grids are finally moving mainstream. Indeed, the opinion that grids are the next major computing architecture, with utility computing being it's killer app, is rapidly gaining momentum in the industry. Naturally, the fact that dozens of teams at different companies are working in parallel to advance the technology also means that there is a great deal of confusion on terminology and concepts, as every team introduces their own by necessity.
Nevertheless, one thing is clear: grid computing is out of the academia, and in the mainstream of computing. We need a comprehensive rewrite of this article to reflect today's state of the technology, architectures, applications and challenges.
As a professional working in this area (I am one of the founders of 3Tera, mentioned above), I have the obvious motivation to invest significant effort in this work. I am also a former physicist who have spent a few years in academia, so I know the difference between original research, review articles and marketing :-). This is a significant undertaking, so I will need help with editing, checking sources, etc., and, of course, with keeping me honest :-)
I would like to hear thoughts from the community here before I embark on this refresh. Maverick61 21:32, 5 August 2007 (UTC)
I propose changing the first bullet point in the article to:
"a collection of independent computers and local clusters combined into a unified computing infrastructure" Jwozniak 20:08, 8 October 2007 (UTC)
I recently sorted the "See also" lists according to blue links, red links, and external links. Donal Fellows pointed out to me that the sorting before was (nearly) alphabetic, which is arguably better. I was confused partly because the alphabetical order had deteriorated somewhat, but I was actually motivated by the recent addition of a commercial link. I was bothered that that might be spam, but I could not easily what the other links were about, especially since the wiki links were indistinguishable from the external links. In the end I removed the three dot.com links, but I also question whether it is useful to have so many red links, or indeed any at all under "See also". I still maintain that a cleanup of the lists is in order, but I certainly don't insist that my way is best. Other opinions? -- Art Carlson 15:24, 8 November 2007 (UTC)
I'm going to contribute by reformatting tables using Help:Sorting for Table feel free to revert Seanwong ( talk) —Preceding undated comment added 01:14, 18 October 2018 (UTC)
Rereading this page, I see that most of the text in it seems to be defining a Grid as doing things like cycle scavenging and stuff like that. That's not what a Grid is about! More traditional supercomputers and Beowulf clusters also fit into the general Grid model, and the interesting use-cases involve coördination of these different resources to solve a problem. For example, using a cycle-scavenging system to check a large set of molecules for whether they look likely to bind to some target protein, and then using HPC to examine in detail the cases that look at least not impossible after the initial scan. (Doing such things often requires resources spread across more than one organization — requiring non-trivial security — and very careful control over QoS so that the coördination works at all...) To my mind, cycle scavenging is just how you build a particular class of (cheap, low interconnect capability, variable availability) cluster. But then I think my view of this is pretty much what Foster and Kesselman's original vision involved. (Grid computing is probably the same as Cloud Computing, and is closely related to SOA.) Donal Fellows ( talk) 23:34, 1 February 2008 (UTC)
CORBA is important for distributed computing in general, but, AFAIK, no Grid middleware is using CORBA at the moment. I thus removed it from the list of Grid APIs. IMHO, Corba is lakking the scalability (mostly in terms of administration/coordination) to be useful for Grid systems. —Preceding unsigned comment added by Andremerzky ( talk • contribs) 10:39, 24 September 2008 (UTC)
ABC@home used to be fast, http://web.archive.org/web/20120520163950/http://boincstats.com/en/stats/41/project/detail should it be added with "as of May 2012"? The numbers I found (might find even higher for this project or another) than the just updated BIONIC and Folding@home I just updated. This would be kind of misleading since the numbers are way lower now (and have been for a while?). Also It's not clear to me if this (or comments like these on talk pages) should go at the top (seems to be almost in that order here). Comp.arch ( talk) 11:58, 23 April 2013 (UTC)
Hello fellow Wikipedians,
I have just added archive links to one external link on
Grid computing. Please take a moment to review
my edit. If necessary, add {{
cbignore}}
after the link to keep me from modifying it. Alternatively, you can add {{
nobots|deny=InternetArchiveBot}}
to keep me off the page altogether. I made the following changes:
When you have finished reviewing my changes, please set the checked parameter below to true to let others know.
This message was posted before February 2018.
After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than
regular verification using the archive tool instructions below. Editors
have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the
RfC before doing mass systematic removals. This message is updated dynamically through the template {{
source check}}
(last update: 5 June 2024).
Cheers. — cyberbot II Talk to my owner:Online 08:06, 28 August 2015 (UTC)
Hello fellow Wikipedians,
I have just modified one external link on Grid computing. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{
Sourcecheck}}
).
This message was posted before February 2018.
After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than
regular verification using the archive tool instructions below. Editors
have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the
RfC before doing mass systematic removals. This message is updated dynamically through the template {{
source check}}
(last update: 5 June 2024).
Cheers.— cyberbot II Talk to my owner:Online 06:15, 30 April 2016 (UTC)
Hello fellow Wikipedians,
I have just modified 6 external links on Grid computing. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.
This message was posted before February 2018.
After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than
regular verification using the archive tool instructions below. Editors
have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the
RfC before doing mass systematic removals. This message is updated dynamically through the template {{
source check}}
(last update: 5 June 2024).
Cheers.— InternetArchiveBot ( Report bug) 05:55, 24 October 2017 (UTC)
An IP has alleged a potential problem. I removed their comment and replaced it with a maintenance tag. Ifnord ( talk) 21:36, 3 June 2020 (UTC)
Heterogeneous computing and grid computing appear to be functionally indistinguishable. Ethanpet113 ( talk) 23:13, 28 August 2020 (UTC)
I am opposed the this merger. Heterogeneous computing is concerned with the micros-scale, within the bounds of the computing node. Grid Computing is on the macro-scale and describes the challenges of federating services across data centers. — Preceding unsigned comment added by 217.140.110.7 ( talk) 19:57, 30 September 2020 (UTC)
I'm also opposed, for the same reason: heterogeneous computing is (simplistically) about stuffing GPGPU cards into servers, whereas Grid is about access to a variety of remote, shared resources. It doesn't help that this article presently fails to mention "virtual organisations" or "single sign-on" which are basically what distinguishes Grid [1] from clouds, clusters and other distributed computing frameworks. -- Louis Knee ( talk) 17:39, 4 November 2020 (UTC)
Absolutely not. Heterogenous computing and Grid computing are different concepts which should not be confused. While a grid system can be heterogenous (with multiple ISAs), it does not have to be. Neither does a heterogenous computing system need to involve multiple simultaneous networked or even interconnected nodes. Heterogeneity can also be expressed over time ( checkpointing, with resumption on a single node later). Johan Hanson ( talk) 16:50, 15 December 2020 (UTC)
References