Posts by Kevint

61) Message boards : Number crunching : Why is it so hard to get work? (Message 23259) Posted 25 May 2009 by Kevint Post: The folks using scripts are perpetrating a DoS attack. The intent of the scripters is to force their own way to the head of the line so that the scripters can have it to themselves, thereby making the service unavailable to others. The attack fits the Wikipedia definition: "an attempt to make a computer resource unavailable to its intended users. ...it generally consists of the concerted efforts of a person or persons to prevent an Internet site or service from functioning efficiently...". See the full Wikipedia article Denial-of-service attack. The article establishes that the scripters are perpetrating a DoS attack. Correct, that is the defintion of DOS attack. And...... We are not experiencing a DOS. You can still get to the web site, in fact, it is much more responsive than other projects I am working on. Since the web pages, message boards and BOINC servers/feeders all use the same IP address, where is this DOS attack? And just who are the intended users here? You? Of course it is. Me? Of course it is. Verstapp - the bad man scripter - Of course it is. Can you stil get to the server? Yes you can. Can I still get to the server? yes I can. So even if your post regarding the definition of a DOS attack may be correct, it is not correct here. The server is not generating any less work because of scripting. And those that are scripting do not have the intent of crashing the servers or the site. There are much better and easier ways to inflict a DOS attack. And "PING" is not one of them.
62) Message boards : Number crunching : Why is it so hard to get work? (Message 23258) Posted 25 May 2009 by Kevint Post: @Bill You have it totally backwards. Lack of work is not caused by those that use scripts - PERIOD. The reason there are those using scripts is because lack of work. Some people here have real short term memories I think. When this project first started, and for months afterward, work was hard to get.. after several months of on and off work, and several fixes to the WU generator, things settled down and work was plentiful, ( then came the GPU. And many of us jumped on the GPU bandwagon. Then as more and more started attaching new hosts, and running more and more GPU's the work became more and more scarce. And this continues, have you noticed that the number of new hosts on daily basis keeps going up, and the amount of work done stays constant? Simple rule of supply and demand. More hosts against a finite amount of work, means less work per host. As more hosts attached, available work per host became less and less. Then someone decided it was a wise and smart thing to announce to the world the use of scripts. It was not the other way around. If work as readily available, scripts would not have been used in the first place. The work generator is the culprit, it just can not keep up with the current work demand. Scripting is not slowing down the work generator. By implementing a back-off routine like you suggested will not produce more work. It will only spread the work out a bit more between hosts, but not much. As those with larger farms will still be requesting more often. The fix is to generate more work, or generate longer WU's for the GPU. Hmmm. Maybe we should have a GPU only project. Never at anytime did anyone promise anyone a full cache of work. If you feel like you are not getting "your share" then attach to a different project or a couple of different projects.
63) Message boards : Number crunching : 4 * GPU Slots! Oh My! (Message 22559) Posted 17 May 2009 by Kevint Post: actually if you guys read the article... it's 7 pci-e 16x 2.0.......... 7 slots..... Yes this is true, but each card takes some width, and looking at the board itself the slots are too narrow to fit that many cards ATM. Each card may only take one PCI slot, but the width of the card overlaps the other slots. If card makers would build the cards so they don't take so much real estate it would be possible to install more than 4. Maybe someday in the future the card makers will figure a way to make them skinny or the board makers will give more space for each PCI slot. I have some older Intel boards that have 5 PCIe slots but can only physically fit 3 cards, not at 16x for each slot. But it does not seem to matter with MW if they are 16x 8x or 4x, at least not from what I can tell.
64) Message boards : Number crunching : 4 * GPU Slots! Oh My! (Message 22542) Posted 17 May 2009 by Kevint Post: I think my R.A.C. started to go up just looking at that MB!!!! ;-p That is not your RAC. This is a very sweet board. Feasibly 4x16 PCIe slots. - loaded with over clocked Dark Knight 4870's 400K per day. Very nice.
65) Message boards : Number crunching : WUs distribution (Message 21758) Posted 9 May 2009 by Kevint Post: Why would he bother with it at all? It only means that much more science is getting done that much faster. But it is getting done on other projects, not his beloved SETI (washes mouth out with soap) IMO the reason DA is so hell bent on cross project parity,(if the project is granting more than SETI, if less than SETI. ohh well) to push more participants back to SETI.
66) Message boards : Number crunching : Milestones (Message 21742) Posted 9 May 2009 by Kevint Post: ...... but I did go over 100,000,000 Total BOINC Myself this week ... :) Nice crunching there PB!! Congrats !!! Looks like I am going to have to invest in some nVidia cards soon as well -
67) Message boards : Number crunching : Milestones (Message 21342) Posted 5 May 2009 by Kevint Post: Lots of zeros 100,000,000 in Milkyway 206,000,000 total BOINC Holding 10 #1 spots 34 top 10 positions #1 BOINCstats World Cup, 2 months running.
68) Message boards : Number crunching : MilkyWay_GPU - Almost There! (Message 21263) Posted 4 May 2009 by Kevint Post: This GPU angle, using CUDA rather than Cluster Physics ATI client, Do you have some inside knowledge that nobody else does? Last I heard they were going to support both GPU apps. The project to develop their own for CUDA, and still support third party CAL.
69) Message boards : Number crunching : MilkyWay_GPU (Message 19854) Posted 21 Apr 2009 by Kevint Post: I know its "outdated" I didn't ask about that. I asked if the project will SUPPORT it. I went with the 8800GT because my mobo looks like it only supports a single pci-e card. I guess the only ones that could tell you would be the programmers that are suppose to be working with Travis to produce the CUDA based app. Currently as I am sure you know, the only GPU card that is works in MW is ATI 3800 and 4800 series. There has been some discussion about a CUDA app being developed but other than that, there as been no promises. So we are just in wait and see mode for now.
70) Message boards : Number crunching : GPU units hanging (Message 18076) Posted 9 Apr 2009 by Kevint Post: Ooh ok. Having a look. Couldn't even find that section using the search tool. I opened it up using notepad and nothing even resembling it. Scratching my head here. This is the complete contents of my cc_config.xml for my quads with GPU cards. On my duo core boxes ncpus = 3, Single core boxes ncpus=2 <cc_config> <options> <ncpus>5</ncpus> </options> </cc_config> Remember you ether have to restart BOINC or in BOINC manager - Advanced Read config file You can make changes to the cc_config file on the fly by re-reading the config file in BOINC manager.
71) Message boards : Number crunching : ATI 4750/4770/4890 support (Message 18075) Posted 9 Apr 2009 by Kevint Post: No, he does not. Misfit, do you ever have anything constructive to say?
72) Message boards : Number crunching : GPU units hanging (Message 18003) Posted 8 Apr 2009 by Kevint Post: Thanks for the advice, will give it a go. This is what mine looks like:- <app_info> that is your app info file, found in the project folder. The cc_config.xml file is a different file. It is found in the same folder as your client_state.xml folder. <cc_config> <options> <ncpus>5</ncpus> </options> </cc_config>
73) Message boards : Number crunching : GPU units hanging (Message 17977) Posted 8 Apr 2009 by Kevint Post: How do you have your <ncpus>xx<\ncpus> in cc_config.xml set? I have seen this problem on several of my boxes, sometimes I have had to just run ATI MW exclusively. For no apparent reason it does like to play well with other projects on some boxes. I have my setting ncpus=cpucores+1 sometimes helps. On other boxes, I can run 1 other WU.. . I choose CPDN because they have long WU's. I believe this is happening because BOINC does not recognize ATI/CAL and has real trouble with figuring out what app to run on what core. I know this is not a lot of help, but it works for me.
74) Message boards : Number crunching : Travis: Please set a minimum update interval (Message 17690) Posted 6 Apr 2009 by Kevint Post: --<snip>-- Either way, setting update limits will not increase the amount of available work. --<snip>-- You're not paying attention. Nope I think it is you that is not paying attention. Ether that, or not understanding what I said. Setting a back off interval is fine in theory, but it will still not make the lack of work situation any better. I said. In order to get more work, get more boxes running. The more boxes running, the more work you get. Plain and simpl Work is being distributed amongst the hosts. If you want more work, get more hosts. WU's are being distributed as fast as they are being generated. And will continue to do this no matter what sort of back-off timer you put in place. The server is simply at its limit. It is very possible, and most likely that the project admins reduced the flow of work after the last crash to keep the servers from crashing again, or to lower the amount of network usage. There were complaints from the school about the high bandwidth usage just before the last crash.
75) Message boards : Number crunching : Travis: Please set a minimum update interval (Message 17672) Posted 5 Apr 2009 by Kevint Post: Please set a minimum update interval According the the graph, and the topic of this thread. Setting a minimum update interval will do absolutely no good. The project has reached its WU production peak. True, its peak output now is much lower than it was before the crash a last week. Before the crash, the project was doing 20M a day in credit, now is is doing 15M a day in credit. Could be 2 causes for this. Lower credit per WU -(I have not checked to verify this, maybe someone should) OR, the server is peaked out, and the project admins have lowered the WU creation to avoid another crash. Either way, setting update limits will not increase the amount of available work. In order to get more work, get more boxes running. The more boxes running, the more work you get. Plain and simple. I don't like running at 50% any more than the rest, but I also run other projects alongside MW to take up the slack.
76) Message boards : Number crunching : Server Updates and Status (Message 17245) Posted 1 Apr 2009 by Kevint Post: We'll be running them as separate projects, with separate top lists and all that. I think this is the best approach so we can have lists of our best CPU and our best GPU crunchers. It will also easily allow users with both GPUs and CPUs to connect to both and (hopefully) run both at the same time -- at least that's our goal. Are you sick of these questions yet ? How are you going to handle the current credit data base? Freeze it and start over with new projects? Or are you going to determine what gets moved to what project? If the later, how are you going to determine what gets moved to what project? Ex, I have approx 25M on CPU, and 25M on GPU.
77) Message boards : Number crunching : Milestones (Message 17233) Posted 1 Apr 2009 by Kevint Post: 50,000,000 for MW!
78) Message boards : Number crunching : Server Updates and Status (Message 17232) Posted 1 Apr 2009 by Kevint Post: Just a heads up on our situation. We should be meeting with labstaff this week to set up the milkyway_gpu project (it'll be at http://milkyway.cs.rpi.edu/milkyway_gpu and perhaps milkyway_gpu.cs.rpi.edu). Once that's up and running I should have a preliminary application and code out there for people to use with it shortly after. Any statement on how the project will run afterwards? Is there going to be 1 project or 2 projects - for those of us that run both CPU and GPU apps.. From the URL change it appears that we will have to attach to a different project? How are stats going to be held, as a single project or 2 projects?
79) Message boards : Number crunching : ATI Crunchers dominate top 100 list (Message 17002) Posted 27 Mar 2009 by Kevint Post: ...buy the crappiest system they have with a PCIx slot... Make that a PCIe slot, and the 4870 will fit ;) (Sorry, couldn't resist.) Corrected. I should post some pics some day of a couple of the boxes I have running 4870's.. if you can call them boxes...
80) Message boards : Number crunching : ATI Crunchers dominate top 100 list (Message 16983) Posted 27 Mar 2009 by Kevint Post: Alternatively, for the short term, Intel Mac OSX users could attempt to run VMware Fusion and see if that gives enough support to run the GPU app... A copy of Windows is required as well as the Fusion application. For the less cost as of all that, you could go to a used computer store, buy the crappiest system they have with a PCIx slot, 40 gig HD, get a used 4870 (I have seen them for as low as $130.00) and walla - 80K a day. All for less than messing around with VM ware, and Fusion - and bloating a GREAT OS (MAC) with something that is as crappy as windoze. Having windows run inside a VM on a MAC is like introducing my ex with my current. Not a wise choice IMO.

Previous 20 · Next 20