Posts by Beyond
log in
1) Message boards : Number crunching : Spliting GoofyxGrid - in progress (Message 1633)
Posted 29 Mar 2017 by Profile Beyond
All of my PCs defaulted to CPU intensive tasks.
Why not keep nci as default and let people add cpu tasks if they want?

Yes, NCI should be the default.

Setting Goofy to no new work until this gets fixed.
2) Message boards : Number crunching : Monkeys - CPU (Message 1565)
Posted 23 Feb 2017 by Profile Beyond
In next 2-3 weekes I will be checking again possibility of turning on CPU apps on my server. I just now want to sorry for any inconvenience about it <- on some host NCI app could block CPU slot itp

I really need to start CPU apps against creating more NCI monkeys app.

This is a fun project to run as an NCI app but bumping other projects to use valuable CPU cores? I wonder how many will really want to do that?
3) Message boards : Number crunching : 'Server Can't Open Database' (Message 1409)
Posted 6 Feb 2017 by Profile Beyond
OVH send me mail that there is to much connection to my server and they restart it in rescue mode :(
Now I restart it in standard mode and turn on project.

One of the problems of allowing multiple host connections...
4) Message boards : Number crunching : High Credit (Message 1278)
Posted 20 Dec 2016 by Profile Beyond
Hi Admin, thanks for the reply and the unbanning of accounts. To assure that anyone who has the wish and the capability to run multiple instances can participate without having the feeling that others are "cheating", it would be a nice idea to configure a setting in the Goofy@home preferences where you can select the amount of concurrent instances you want to run.\r\nTo prevent errors, this could be capped at max 150 instances per machine.\r\n\r\nThis would give the project a very nice boost and will terminate the accusations related to unfair use of VM's / running multiple instances outside the BOINC app, etc.
5) Message boards : Number crunching : Testing NCI i CPU (Message 1066)
Posted 25 Oct 2016 by Profile Beyond
Why not set up 2 separate BOINC servers, one for NCI apps and one for CPU apps?
6) Message boards : Number crunching : High Credit (Message 1016)
Posted 11 Oct 2016 by Profile Beyond
Actually, that only shows that they are running an excess number. The rules should apply to all including those just running 1 or 2 extra instances. At least until there are actual posted rules indicating a limit. I've known people with data centers that had a lot of hosts identical. So, until they go into the thousands, it can be quite difficult for some to be proven by that list.
\r\nIf you take all of the data points I listed, it's not too difficult to see many who're running multiple instances. Like we both say, we need clarification of the rules, if any.\r\n\r\nSomehow I feel like I'm not being understood. I haven't mentioned "cheating" once. How can there be "cheating" if there's no rules? What I keep asking for is a clarification of the rules. If multiple instances are OK, then I'd like to run some too. But I won't unless it's OK with the project Admin.
7) Message boards : Number crunching : High Credit (Message 1014)
Posted 11 Oct 2016 by Profile Beyond
How do you enforce multiple instances rule? How do you prove how many hosts they have?
\r\nA problem but there are some strong hints:\r\n\r\n1) Many, many machines with the same BOINC specs.\r\n2) Machines with the exact same benchmarks.\r\n3) Machines with strange benchmarks such as: \r\n- Measured floating point speed 1000 million ops/sec\r\n- Measured integer speed 1000 million ops/sec\r\n(No real machine has specs like that.)\r\n4) Hidden machines is another hint, although not a strong one.\r\n5) The Admin has additional ways.\r\n\r\nThis is really kind of academic though since the project admin hasn't set firm rules that are still posted...\r\n\r\n
The lack of rules governing how the donors are allowed to contribute is a bigger problem.
8) Message boards : Number crunching : High Credit (Message 1012)
Posted 11 Oct 2016 by Profile Beyond
The setting FreeHAL had would apply to all hosts and therefore people could still run multiple instances and get an insane amount.
\r\nThere wouldn't be any point if it was unlimited. If it was limited, have a hard rule against multiple instances and enforce it. \r\nThat way the Admin could also adjust his desired project WU level by simply changing the max allowed (in the project preferences) from time to time.\r\n\r\n
Like other suggestions...probably best to release a CPU intensive version and then adjust credits according to the amount of work being done.
\r\nAs you say, probably the best idea, but of course it's up to the Admin.
9) Message boards : Number crunching : High Credit (Message 1010)
Posted 11 Oct 2016 by Profile Beyond
If the Admin wants to allow multiple instances it's his decision. The reasonable way to do that would be to set the number of instances in the project preferences. As I remember that's what FreeHal did (with a limit of 15 in that case). Personally I feel sorry for the Admin. All he wants to do is run a nice little project that he devised, and then it gets put in the highly competitive DC Vault and all heck breaks loose. I know of at least 3 ways to do a large number of multiple instances. The question is: is that what the Admin wants? I won't do it without his permission.
10) Message boards : Number crunching : High Credit (Message 1008)
Posted 11 Oct 2016 by Profile Beyond
Milkyway and Poem were also popular projects for running more than one instance (again 2-4). That was because the code was not efficient enough to optimize the GPU use. It was fine with the admins. The project I run the most is GPUGrid and a few people with very fast GPUs run 2x WUs. It's fine with the admins even though they prefer 1x in order to minimize turn around time. None of these projects gain a huge amount by running multiple instances and it's only done to maximize the valuable GPU resources. Very different than an NCI project where almost limitless copies can be run with little effect on resources.\r\n\r\nFirst we need a clear statement from the Admin. There's also a very easy solution to this whole question:\r\n\r\nIf multiple sessions are going to be allowed: just implement a setting in the project config to allow people to run as many instances as they like.
11) Message boards : Number crunching : High Credit (Message 1005)
Posted 11 Oct 2016 by Profile Beyond
Cheating occurs when there is a clear rule that is being broken. The current unspoken rule is 1 of each application per host. They ran just 2 per host. Host (in BOINC terms) is not defined as a physical machine. I completely disagree that the masses will say it was unfair.
\r\nI wish that the Admin would answer the question that I asked above. A 1 minute search yielded this posted on several team forums:\r\n\r\n"Please don't create special virtual machines for my project <- in my opinion is not fair in relation to other. For now i had sent 3 warnigs message to user to abandon that way of crunching.\r\nI don't want to have situation like was with FreeHal project. 2 days ago I had to ban and delete 3 user account with more than 150 VM's.\r\nToday I sent next warnings to 3 users."\r\n\r\nSeems clear enough, but now it appears that whole thread has been deleted.\r\n\r\n
Is running GPU's with app_configs allowing multiple work units per card unfair at GPU projects? Think long and hard before answering.
\r\nNot the same. Goofy uses virtually no resources so it should be possible to write a script to spawn a vast number of sessions with little impact on the machine. It was intentionally designed to be an NCI app. App configs for GPUs simply allow a few (generally 2-4) WUs to run simultaneously for apps that are not completely utilizing the GPU. AFAIK this has always been allowed and even encouraged by the Admins.\r\n\r\nIf multiple sessions are going to be allowed, why not just implement a setting in the project config to allow people to run as many instances as they like?
12) Message boards : Number crunching : High Credit (Message 996)
Posted 6 Oct 2016 by Profile Beyond
Since december 2015 I blocked 10 users who used boinc multiple instances <- with massive count. For example 3 users had more that 10000 instances each.\r\nI know my project is NCI but even that there is a border for cheating.
\r\nPretty much all the top users are finding various ways to rack up huge numbers in order to "game" the DC Vault scores (multiple BOINC instances, config hacks, virtual machines). Some teams have multiple users doing this. If you want this to happen let everyone know and bar the doors. If you don't want it to happen, sanction the offenders and pull their scores. It's really your decision to make. Just let us know what you decide.

Main page · Your account · Message boards

Copyright © 2019 Goofyx

Generated 22 Oct 2019, 16:33:57 UTC