Message boards :
Number crunching :
CNODE Runtime
Message board moderation
Author | Message |
---|---|
![]() Send message Joined: 22 May 20 Posts: 11 Credit: 1,711,704 RAC: 190 ![]() ![]() |
While PERF WU's take just over a couple of minutes to earn their 2 credits, CNODE WU's run for up to a whole day for their 2 credits. How long do you expect support for CNODE ? |
![]() Send message Joined: 22 May 20 Posts: 11 Credit: 1,711,704 RAC: 190 ![]() ![]() |
Many CNode tasks take ½-1hour. Mine take a day or more - and often end with an error. |
Send message Joined: 12 Nov 19 Posts: 34 Credit: 115,884,814 RAC: 201 ![]() ![]() ![]() |
My Linux CNODE tasks almost always run right around 30 minutes unless there is a problem. That is a lot longer than the Windows PERF tasks run, but it has pretty much always been that way. The two applications run different types of tests. I'll be honest that I don't want to look through all of your computer list/task history to get a more detailed report, but from what I did see, it looks like you're running Linux in a virtual machine on a Windows 10 laptop with an i5-6200U processor. I see that the VM used to be allowed two CPU cores until last October, and now the new(?) VM is only allowed one core, and apparently a severely restricted one based on the extremely low BOINC benchmark score. A side result of that is you are actually having some tasks run as long as 42-48 hours before they are able to complete their assigned work. That's definitely not normal. While iThena is technically an NCI (Non CPU Intensive) project, running larger numbers of instances of BOINC/iThena at the same time do add up and require a modest amount of CPU power. My guess is that the 40 or so instances you are running in that VM need more CPU power than what you are letting the VM have. I suspect you'd get much better results if you let the VM actually have a full CPU core (or even half of a core) instead of just a small percentage of one. That would potentially reduce the productivity of the laptop in other projects a little, if you are running any, so you'd have to decide if running Linux iThena tasks is worthwhile to you or not. The CNODE tasks also use a LOT more network bandwidth than the PERF tasks, so if your Internet connection is limited, running multiple Linux CNODE tasks could be overwhelming your connection. Personally, I choose not to really focus on Linux and just let any of my actual Linux machines run a single iThena instance. They tend to average about 30-50 credits per day, while my Windows systems get 150-200 or more credits each per day. |
![]() Send message Joined: 22 May 20 Posts: 11 Credit: 1,711,704 RAC: 190 ![]() ![]() |
Thanks @Fardringle for your analysis - it makes a lot of sense. The VM has a whole thread, but doesn't appear to use it all - except when applying Linux updates.- which is why I scaled down from 2. Net activity is low - rarely exceeding 100Kb. I'll go back to more threads - the machine isn't doing much else now (about 20 PERFs which run imperceptibly) - and it's probably resource conflict that lies at the "core". Especially when all resources have to be funneled through VM. 40 parallel CNODEs is probably far too optimistic... This is my first experimental venture into the world of Linux - I'm thankful for all the help out there. |
Send message Joined: 12 Nov 19 Posts: 34 Credit: 115,884,814 RAC: 201 ![]() ![]() ![]() |
Frankly, unless you specifically want a Linux "badge" from the project, or you just want to do it while learning more about Linux, I would just run iThena in Windows. The Windows tasks run a lot faster, so they get a lot more credit, and they don't use anywhere near as much Internet bandwidth. |
![]() ![]() Send message Joined: 24 Sep 19 Posts: 111 Credit: 1,004,738 RAC: 1,171 ![]() ![]() ![]() |
Since the downtime the new work units are doing what they were doing back in March. They are running for hours. I have a CNODE that is at 22% after 55 minutes. Also Windows PERF is running for 30 minutes or more, I have one at 99.996% after 27 minutes. I managed to get a rare PERF for Linux to download and it is also running as long as the Windows PERF work units do. You said you would think about doing a credit bonus to the users that had work units running for many hours back in March. That did not happen, so can we get a double bonus this time? Conan |
Send message Joined: 12 Nov 19 Posts: 34 Credit: 115,884,814 RAC: 201 ![]() ![]() ![]() |
Well, my Linux CNODE tasks were taking about 30 minutes per task back when this was originally posted. I honestly don't pay a lot of attention to them so I don't know when it changed, but all of my CNODE tasks are taking about 24 hours to run now, just as other people mentioned in this thread. 30 minutes was bad enough, compared to the 3 minutes or so that the Windows tasks run, but 24 hours is just awful. Virtually zero credits, and they still run all day and use up Internet bandwidth all day. I'm turning my Linux instances off until this gets resolved. |
© 2019-2025 iThena. All rights reserved. | Private Policy