 |
|
|
1)
Message boards :
Number crunching :
New work discussion - 2
(Message 66306)
Posted 9 Nov 2022 by Jim1348 Post: These modles are still in testing. I am setting up several machines and swapping memory between them for an optimum allocation. I think I will wait until next Spring when you have it all worked out. |
|
2)
Message boards :
Number crunching :
New work discussion - 2
(Message 66301)
Posted 8 Nov 2022 by Jim1348 Post: Some info on memory requirements on upcoming OpenIFS forecasts. As mentioned previously, we're aiming to increase the model resolution to be more scientifically valuable. These resolutions come with higher memory requirements: https://www.cpdn.org/forum_thread.php?id=9149&postid=66077#66077 I don't recall seeing any discussion on how often the peak is reached, or what happens when a machine's memory is exceeded. For example, if you have 64 GB, will BOINC send you the right assortment of work units to make best use of that, or do you just wait until more memory is available? And if the peak is not reached very often, can you run more than two of the 24 GB work units at a time? |
|
3)
Message boards :
Number crunching :
New work discussion - 2
(Message 66283)
Posted 2 Nov 2022 by Jim1348 Post: The AMD Ryzens with a large cache-to-core ratio have always worked well for me on the CPDN projects. That is why I am using Ryzen 3600's. The have 32MB of total L3 cache and only 12 virtual cores. I often restrict that to 50% of the cores, so the work units are usually running on full cores in effect. I have heard that OpenIFS does not take so much cache, so I may be able to use 75% of the cores. |
|
4)
Message boards :
Number crunching :
New work discussion - 2
(Message 66271)
Posted 29 Oct 2022 by Jim1348 Post: They could also just allow the user to set the maximum number of work units running, or at least downloaded at a time. LHC does the former, and WCG does the latter. |
|
5)
Message boards :
Number crunching :
New work discussion - 2
(Message 66263)
Posted 28 Oct 2022 by Jim1348 Post: "BOINC starts multiple OpenIFS tasks because there are free CPU slots, even though the total memory for the tasks exceeds what's available. " As I understand it, they are going to run two cores per work unit at first, so you will have only two work units running. The memory should be enough. (But yes, if you limit the number of BOINC CPU cores, then that will limit the number of work units running.) |
|
6)
Message boards :
Number crunching :
New work discussion - 2
(Message 66260)
Posted 28 Oct 2022 by Jim1348 Post: Assuming that also includes linux in virtualbox and WSL, when upwards of ~5000 linux openifs tasks go out, that's plenty of work. Some of my work will require the higher memory machines. Very good, I can put two Ryzen 3600's on it, one with 64 GB and the other 128 GB. But they may get into a fight over bandwidth across the Atlantic, which seems to be limited to 10 Mbps for me. I may have to back off to one machine. |
|
7)
Message boards :
Number crunching :
New work discussion - 2
(Message 66258)
Posted 27 Oct 2022 by Jim1348 Post: I am inclined to think that they will go quickly too, though the limits are probably more on bandwidth than memory for most people I think. But once the word gets out, there will be a lot of people willing to try at least. |
|
8)
Message boards :
Number crunching :
Windows & WSL give quite different CPU benchmark results in boinc?
(Message 66251)
Posted 26 Oct 2022 by Jim1348 Post: I run WSL2 with boinc, though it does still freeze on me now and again. I get the freezes too every couple of weeks, so I don't use it any more. I hope they fix it with an upgrade. Maybe it works better on Win11, but I would prefer to avoid that as long as possible. |
|
9)
Message boards :
Number crunching :
Windows & WSL give quite different CPU benchmark results in boinc?
(Message 66248)
Posted 26 Oct 2022 by Jim1348 Post: I have done a bit of Win10 WSL2 work on various projects, and find that it runs just as fast as my Ubuntu 20.04 machines on comparable hardware. I wouldn't pay much attention to any BOINC measurement. https://www.cpdn.org/forum_thread.php?id=9025#63462 |
|
10)
Message boards :
Number crunching :
New work discussion - 2
(Message 66215)
Posted 22 Oct 2022 by Jim1348 Post: Trying to get two different Linux services running together would be a different kettle of fish. I can send you my sample files, and explain why the various elements are needed, if it would help.Are you referring to two BOINC instances? I do it all the time on Ubuntu. I find it easier to set up than on Windows, or at least to start up both instances. https://www.overclock.net/threads/guide-setting-up-multiple-boinc-instances.1628924/ Then, I use BoincTasks to manage both instances (on different ports) on Ubuntu machines from my main Windows machine. I have used it when one BOINC instance is on CPDN, and the other instance on another project. I have not tried both instances on CPDN, but it probably will work. (I had a problem on one project a few years ago when both instances were on the same project, but not recently. I think the current BOINC server version works OK.) |
|
11)
Message boards :
climateprediction.net Science :
Climate change in the News
(Message 66214)
Posted 22 Oct 2022 by Jim1348 Post: As OpenIFS doesn't have an ocean model, only analysed sea-surface temperature fields, the model's atmosphere doesn't get that feedback. If it is just a question of memory, let us try the whole thing. I have 128GB waiting to be used on one machine, and could do two easily enough. (But I get the impression that it is not quite so simple.) |
|
12)
Message boards :
climateprediction.net Science :
Climate change in the News
(Message 66210)
Posted 22 Oct 2022 by Jim1348 Post: Thanks, I think the new study just updated it to show the accelerated heating in the last 10 years. But doesn't OpenIFS still predict storms? If not, what are you doing with it? I have not seen a description for a while, but maybe I missed it. |
|
13)
Message boards :
climateprediction.net Science :
Climate change in the News
(Message 66199)
Posted 20 Oct 2022 by Jim1348 Post: Oceans are warming faster than ever. Here’s what could come next. https://wapo.st/3Deg6zK I trust that OpenIFS will help to predict the effects of the larger and more powerful hurricanes. |
|
14)
Message boards :
Number crunching :
New work discussion - 2
(Message 66196)
Posted 18 Oct 2022 by Jim1348 Post: Looks like a batch of HADSM4'S are being prepared. I just picked up a couple, and they are _0. |
|
15)
Questions and Answers :
Unix/Linux :
*** Running 32bit CPDN from 64bit Linux - Discussion ***
(Message 66105)
Posted 14 Sep 2022 by Jim1348 Post: Have you noticed it with different computers?I have only one Win10 machine, but it has been through several updates (also updates on the Linux side), and the problem still persists for over a year. https://www.cpdn.org/forum_thread.php?id=9025&postid=63462#63462 The machine is stable otherwise. |
|
16)
Questions and Answers :
Unix/Linux :
*** Running 32bit CPDN from 64bit Linux - Discussion ***
(Message 66102)
Posted 13 Sep 2022 by Jim1348 Post: I believe there are at least a few other WSL2 users here too. Maybe it is your hacking. :-) WSL2 works great for me with CPDN (Ubuntu 20.04), except that after a couple of weeks it just stops and I have to reboot. I think others have noted the same problem. But I run my machines 24/7. If you reboot anyway, you probably will not notice that problem. However, you then have the problem that CPDN will not always survive a reboot. You really should pause it before rebooting, and it should work. |
|
17)
Message boards :
Number crunching :
Hardware requirements for upcoming models
(Message 66023)
Posted 31 Aug 2022 by Jim1348 Post: I wonder if Boinc has knowledge of your internet speed, and could send different types of task to different people? If not, there could always be a tickbox in preferences to accept enormous tasks. We have asked for the ability to select the type of tasks we want for a long time. It seems to be against their ideology, which is always the most difficult to argue against. But I for one won't attempt to mix enormous tasks with ordinary ones. They either select them based on what my machine can do, or allow me to select them. |
|
18)
Message boards :
Number crunching :
New work Discussion
(Message 65996)
Posted 25 Aug 2022 by Jim1348 Post: My usual problem is dropped connections on short files - they can't handle the concurrency. I changed these in my cc_conif.xml: <max_file_xfers>1</max_file_xfers> <max_file_xfers_per_project>1</max_file_xfers_per_project> They are normally set to 8 and 4 respectively. It fixed my last 7 stuck downloads. Thanks for the tip. EDIT: "2" seems to work OK also. |
|
19)
Message boards :
Number crunching :
New work Discussion
(Message 65978)
Posted 24 Aug 2022 by Jim1348 Post: I expect that a lot of us have enough RAM, bandwidth, cores, etc. The tendency of most BOINC projects is to cater to the minimalist user. I see no reason to do that here. Use as many resources as you need and see what happens. It is either enough to be scientifically useful, or it isn't. But you won't know if you limit yourself unnecessarily. |
|
20)
Message boards :
Number crunching :
Hardware requirements for upcoming models
(Message 65960)
Posted 23 Aug 2022 by Jim1348 Post: If the work units are of the same size as the ones that have been through testing, the final uploads have been up to 1.5GB. If for example, tasks increase in size by a factor of four so that they are still completed in the same time running on more cores, that could see uploads of over 6GB the tasks don't send up data continuously while they run. How long do the work units run? I don't recall seeing that. My cable modem is rated at 15 Mbps upload, but in practice I see 20 Mbps. That should upload 6GB (48Gb) in 40 minutes if the other infrastructure supports it. But that is probably the real limitation, not the local speed. I see very good speeds across the Atlantic, but never more than 10 Mbps. So in practice, it would take more like 1 1/2 hours. And with uploads from my other machines, it could be longer than that. |
©2024 climateprediction.net