Talk:Berkeley Open Infrastructure for Network Computing/Archive 1
|This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.|
- 1 Contributors
- 2 Open
- 3 Criticisms?
- 4 Active hosts and TFLOPs
- 5 projects
- 6 External links
- 7 Definition of Active Hosts
- 8 RV of ceea edits,i'm puting it here
- 9 Is MacOS X special?
- 10 Linkfarm
- 11 new name?
- 12 Bossa Info?
- 13 Copied by Knol?
Removed this line from the bottom of the main section:
"In December 2003, Sun Microsystems announced it would donate some of its own products—including Solaris servers, and workstations—to BOINC (Vance, 2003)."
There are stacks of contributors for BOINC. Although Sun`s sponsorship/contribution would be significant, so is the donation of computer power and equipment by thousands of others. BOINC is based on, and completely reliant upon such donations, not only from Sun.
--Antwelm 10:25, 2005 May 24 (UTC)
- Perhaps a section named "Contributors" or "Contributions?" - Jim 10:05, 29 September 2006 (UTC)
Other than the title, the title sentence, and the external link, no where else is the term 'open' mentioned, let alone explained. Is this open as in free and open source? If so, I think that it is very important to mention that in the article. I'll do some research, and since PlanetQuest uses this platform, it would be open as well. --ShaunMacPherson 19:45, 31 May 2005 (UTC)
- BOINC is licensed under the LGPL. Direct quote from : "BOINC is free software, distributed under the Lesser GNU Public License (LGPL)." However, I remember reading somewhere that it does not apply to client software: for example, Climateprediction.net is proprietary. Wikiacc 20:02, Jun 1, 2005 (UTC)
- The entire BOINC software inferstructure is completely open and free, this extends to the common server daemons and general installation package, as well as the BOINC client.
- What is usually not open & free are the science applications that perform the actual calculations, as these are project-specific things.
I think there may need to be a criticisms section in this article. Although I'm a fan of distributed computing and enthusiastic about BOINC, the current implementation isn't quite what it could be. I'm only posting this in the discussion section because I don't want to turn the article into a "What's wrong with BOINC rant". Among some of my criticisms:
- Getting BOINC running on a new computer can be difficult. Though the latest version appears to address this, I was initially frustrated by not having something as clean and simple as Folding@Home and the original SETI@Home client. For those, I can/could just install, enter my E-mail and I was done. This set-and-forget simplicity didn't carry over to BOINC like it should have.
- Web preferences. This is cumbersome and a pain to use. While there certainly may be some advantages, I really should be able to set prefs from within the application like most software.
- Overdone graphics. Though the graphics can be whatever the developers choose, the BOINC version of SETI@Home has no option to go back to the classic look and instead there's that frantic OpenGL graphic.
- That's to complain about SETI, not about BOINC. 188.8.131.52 22:00, 21 February 2007 (UTC)
- No simultaneous computation of multiple projects. When I first heard that BOINC would allow for multiple projects, I imagined running a couple or a few in separate windows as independent threads, like some busy computer in a sci-fi film. I was disappointed when I learned that wasn't possible yet. From a computation standpoint, this may not be as efficient as switching back and forth between them, but BOINC isn't a supercomputer to be fine tuned for maximum performance. It is - or should be - a fun application that can entertain users and happens to be used for computation as well. Since BOINC uses spare CPU cycles, something as simple as animated GIF's and Flash animation ads in a left-open web browser can use as much or more CPU than BOINC. Given the wastefulness of real-world performance, the CPU hit from allowing multiple simultaneous threads appears to be of less importance than keeping users like me happy and hence more likely to use BOINC in the first place. Restrictions of use that make the application less fun are penny-wise and pound-foolish.
- Having more apps running in the background will make your normal tasks (like browser) slower. Recent versions of BOINC have a hidden setting to make it behave as if you had N CPUs (running N tasks at a time). It's hidden because it's for debugging (to test BOINC behaving correctly with multiple apps, when the developer doesn't have a multi-CPU system). 184.108.40.206 22:00, 21 February 2007 (UTC)
Looking at all these problems, I wondered if BOINC may suffer from a self-fulfilling prophecy - If BOINC is only designed for geeks to use, then only geeks will use it. Whether you're a programmer or computer neophyte makes little difference to the performance your computer can deliver. The developers of BOINC need to go back to the drawing board and release a nice, simple, straightforward, and more fun version that includes an easy to use setup wizard. They should ask themselves "Can Grandma use it?" and "Will little kids want to use this?" In addition to a continuation of the existing application, make a version of BOINC that far more people will want to run. Make it attractive enough to go mainstream, get it preinstalled on OS's, and attract more funding so BOINC projects aren't just running on thousands of machines, but millions. Web, print, TV, and radio advertising aren't even out of the question. The true potential for distributed computing has barely been tapped - China has virtually none and there's the upcoming $100 laptop program - and I'd hate for DC through BOINC to languish as something only the very technically inclined bother using.
- The problem with getting this into the article is that you will need to find independent and verifiable sources to back you, otherwise it is disqualified as "original research". --David Woolley 20:42, 19 November 2005 (UTC)
- I admit, to an extent this was a way for me to get my own personal criticisms out there without having to sign up for a forum (and Wikipedia is a reliable archive.), but much of what I said has been mentioned before in various forums and other on-line sources. Read up on BOINC discussions in forums and you'll see what I mean. As solid as the underlying software is for BOINC, signing up to projects and the current GUI are not friendly enough to attract mainstream users like it should. What BOINC needs is to have ordinary computer users who might not know or even care at all about the underlying science install the application because they think it's fun and looks cool. They already do that with fairly CPU intensive screen savers, so the research is a great bonus. I know nothing about programming, but I can easily imagine what a nice friendly BOINC setup Wizard and application could look like and attract millions more users if done right.
- While obviously my entire rant has no place in the actual article, it seems like leaving out any mention of the criticisms that do exist is too much of an omission. Is there a 100% transition from SETI@Home users to BOINC users? If not, then why not? BOINC's current problems seem like an elephant in the room.
Active hosts and TFLOPs
The stats on boincstats were distorted with retroactive credit awarding by CPDN/BBC Climate Change. This awarded credits to long inactive hosts and brought the active hosts up by about 100.000 contrary to factual active host. The same is true for the TFLOPs. The TFLOPs have already been corrected to a little over 400 and the active hosts will drop by about 100.000 in a few days. Until then I will not correct the entry but one should know that BOINC never had 480.000 active hosts or a throughput of about 600 TFLOPs. Tralala 18:03, 4 October 2006 (UTC)
I suppose BOINC Alpha, SETI@home BETA, CPDN Alpha, LHC@Home Alpha don't count? Also if anyone doesn't believe that there was a project called ALife@Home, just Google for it. It has, however, disappeared without a trace credit-wise and website-wise.
- The list of projects is getting long rather. Would it be better if it were moved to another article to make this one more readable? - Jim 10:09, 29 September 2006 (UTC)
- I think that's probably a sensible idea -- Lee Carré 07:17, 21 November 2006 (UTC)
- Further thoughts: obviously have a "main article:" template linking to something like "BOINC Projects", but leave only the main projects on the BOINC page; that is, projects that aren't testing/alpha/beta etc. only the in-production ("gamma testing") projects ;) -- Lee Carré 07:26, 21 November 2006 (UTC)
Unofficial optimized BOINC clients
There has been considerable controversy on various project forms regarding the use of "Optimized" BOINC clients. Without mention of the controversy in the article, I do not think it is fair to include a links to an "Optimized" client. Also, because it's not officially supported, I question if it belongs here without mention in the main article to unofficial software. - Jim 21:48, 23 September 2006 (UTC)
- I agree. The need for optimized BOINC apps has vanished with Seti switching to a work-unit-based and fixed credit scheme. As the optimized BOINC apps were all made for the use combined with optimized seti apps there is no longer a reason to use an optimized BOINC app. Since the windows-link is down and the linux-link was not updated since May 2005 I've removed both links. Tralala 17:58, 4 October 2006 (UTC)
- The point for optimized science apps is getting more work done in less time - helps the science. However, using an optimized BOINC client only helps in credit cheating. Saying "SETI uses fixed credits" is plain stupid, BOINC isn't just SETI, there are other projects. Optimized BOINC app with a project that doesn't have fixed credits (ie. most) will cause inflated credits. 220.127.116.11 22:00, 21 February 2007 (UTC)
Definition of Active Hosts
Removed the following text. These paragraphs derive from a misunderstanding of the statistics displayed on the BOINCStats site. The number of hosts listed as active is accurate. An active host is one that has returned work in the last 30 days. The fraction of active hosts shown on the pie chart is the fraction (29.3%) of total hosts (1.4M) that were active (420k), not fraction of active hosts that were active.
- BOINC is not as widely used as raw tallies of "active hosts" might suggest. One needs to account for the number that are idle and are not processing data. In checks of the BOINC site in fall of 2006, about three-quarters of hosts listed as "active" had done no work for over a month.[verification needed] It would seem that the vast majority of BOINC installations either sat idle or had been disabled or deleted but were nonetheless listed as "active hosts".
- Investigators hoping to use this resource for number crunching need to allow for the actual work that can be produced. "Active hosts" figures do not accurately reflect the amount of computing power available since it includes a large proportion of "active hosts" which are only called active by virtue of being on an "active hosts" list. Most are not active at all.
SETIGuy 18:48, 27 October 2006 (UTC)
RV of ceea edits,i'm puting it here
SETI@home origins of the BOINC platform
SETI@home is one of the better known projects that utilizes the BOINC distributed platform.
The success of SETI@home—which after its launch in 1999 quickly became the most powerful computing network ever assembled — made it clear that distributed computing could be used for many other computing-intensive scientific projects. This original success was accomplished using a non-BOINC software product.
This success was tainted by problems the original software presented. Overly-competitive users were detected faking large amounts of work to gain recognition in the SETI@home community. The original software allowed users to spoof their statistics using a number of methods.
Resolving this scorekeeping problem is one of the reasons BOINC was created.
BOINC as a successor to the original SETI@home software
BOINC software has a great deal of competitive scorekeeping built into it. Competitive scorekeeping aspects of this software risks overshadowing its use for scientific number crunching.
As of fall 2006, the lack of alternate software and alternate networks of users meant that the results of science research using BOINC could not easily be duplicated without using BOINC software and the BOINC online network of users.
Therefore, it could not be readily determined by comparison to another software/user-network combination to what extent this particular software/user-network combination may have affected results. The problems with spoofing results is already well known from the original SETI@home project and it is from this same competitive gameplaying subculture that BOINC arose.
BOINC is not as widely used as raw tallies of "active hosts" might suggest. One needs to account for the number that are idle and are not processing data. In checks of the BOINC site in fall of 2006, about three-quarters of hosts listed as "active" had done no work for over a month.[verification needed] It would seem that the vast majority of BOINC installations either sat idle or had been disabled or deleted but were nonetheless listed as "active hosts".
Investigators hoping to use this resource for number crunching need to allow for the actual work that can be produced. "Active hosts" figures do not accurately reflect the amount of computing power available since it includes a large proportion of "active hosts" which are only called active by virtue of being on an "active hosts" list. Most are not active at all.
BOINC as software in 2006
There are numerous problems with the client software.
There are disclaimers warning about CPU overheating that require the end user be a skilled hardware and software specialist to avoid damaging the equipment upon which this software runs. These warnings can be be seen at the BOINC homepage and are in the "rules" pages at SETI@home.
It is also difficult to install for many users. There is no real help resource beyond that provided by amateur volunteers whose level of expertise is unknown and doubtful. Thousands of help requests are documented in help forums and the problems reported do not fall into any few easily defined categories. [verification needed]
There are mentions in Infoworld of serious backdoor security leak potential. This is due to the use of Remote Procedure Calls as part of the software. RPCs allow remote access to a computer.
Communication with the server is even required to make trivial local adjustments in client preferences. For example, the user must logon to the server to adjust the screensaver display unless the user patches their copy of the client software with manually inserted code.
The physical overheating and network security dangers are sufficient to recommend against BOINC being deployed in enterprise computing environments. The amount of tech support intervention required to make it operate also argue against its use in any but an experimental environment. It is best installed, if at all, on a spare, expendable computer dedicated to this sort of experimentation.
To be considered ready for general distribution both server and client must easily install and operate with minimal technical support since they are intended for use by communities of users who are not IT specialists.
BOINC is not yet ready for widespread distribution because it is experimental, beta-quality software.
- --Pixel ;-) 20:46, 28 October 2006 (UTC)
Criticisms of BOINC
On 00:19, 3 November 2006 user Kencf deleted following section with status as "Criticisms of BOINC - Deleted: POV":
"Criticisms of BOINC
At first look it would be nice to have an universal infrastructure for all distributed computing projects out there, but looks can be deceiving. Main negative issue with BOINC is not even BOINC itself, but the userbase running the BOINC based clients - they are not overly loyal to their project and tend to split up the resources between the various projects. From user side it may be natural thing to do, but if to look at it from the project side then it is not so good to have feature anymore - the WU completion time will be much different than when the project client is the only one running. Because of this the projects with components of serial at nature (next WU(s) will depend on previous WU(s)) may be hard to convince to drop their own client infrastructure and to get into the BOINC bandwagon."
Criticism is always a POV. Why was there the "Criticisms of BOINC" section if it must remain empty?
Does BOINC need no criticism or does it have no criticism or it can not take any criticism?
The badly written criticism simply had no cites. Kencf 22:26, 3 November 2006 (UTC)
Is MacOS X special?
Under Design and Structure / Client Structure you can read that "In Mac OS X, the program is able to dynamically take up extra processor speed while you work, varying how much processor time BOINC receives based on how intensively the computer is being used."
As far as I know, BOINC can do this on all multi-tasking capable platforms (by running at low priority). Is the MacOS X client special in some way or should this bullet point be removed?
What external links were the concern that caused the linkfarm tag to be added? I see primarily Wikilinks in the projects section. Dstumme 14:50, 14 November 2007 (UTC)
- The numerous links to the projects that happen to use BOINC seem to be a larger part of the problem. The article is supposed to be about BOINC platform itself, but it is becoming increasingly dominated by applications that use it. (For example, the Linux article does not specifically try to enumerate all of the the applications it can run.) What about just replacing the entire "Projects using BOINC Framework" section with a link to the existing article section List of distributed computing projects#Berkeley Open Infrastructure for Network Computing (BOINC)? Having the list maintained in two places just encourages them to become de-synchronized and unmaintained. -- Bovineone 21:16, 14 November 2007 (UTC)
- With no objections, I've moved and combined the list into List of distributed computing projects#Berkeley Open Infrastructure for Network Computing (BOINC). That resulting list can now serve as a much more comprehensive directory. -- Bovineone (talk) 20:34, 9 December 2007 (UTC)
Behavior On InterNet Computing? http://boinc.ssl.berkeley.edu/dev/forum_thread.php?id=1690
Boinc has been renamed. The acronym remains unchanged but as from 1 April 2007 denotes Behavior On InterNet Computing. Mo.V Secretary to the Board of Directors BOINC Berkeley 1 April 2007
- I think that is an instance of a April Fools' Day joke. There are no other google references to that name change. -- Bovineone (talk) 19:08, 27 November 2007 (UTC)
Copied by Knol?
In case anyone is interested, after looking at Wikipedia:Mirrors and forks/Knol, I checked the knol article on this subject and verified it is an exact copy of this article's lead. The Knol article was written after the July wikipedia version. -84user (talk) 14:21, 5 August 2008 (UTC)