|
1)
Message boards :
Number crunching :
Project has no tasks available
(Message 5455)
Posted 16 May 2022 by [AF>Libristes] ElGuillermo Post: Yep. That what I mean by "I hope the lessons will be learned and solutions will be found to avoid frustrating people". These unintentionnal massive DDoS are relatively new (at least on "solid" projects like Universe@home) They may be related to the increasing number of mutlicore CPUs in PC. Now, even a simple office desktop has at least 4 cores, and we all know we all have 8+ cores hyperthreaded beasts at home... Only way to get through is, effectively, limiting the competition time period, as project owners certainly won't get enough I/O speed to sustain a massive charge for only two weeks a year. One may imagine some kind of cloud based server, using shared ressources that could be resized according to the actual work, but this would also get increased annual fees :/ I'm pretty sure we'll find a way to play nicely. We have plenty of qualified people to think about it among Boinc enthusiasts :) |
|
2)
Message boards :
Number crunching :
Project has no tasks available
(Message 5426)
Posted 14 May 2022 by [AF>Libristes] ElGuillermo Post: Hi ! Your point is valid : actually, newcomers will have some hard time with the project. Believe me (and I am pretty sure I talk for almost all the people involved in this "competition stupidity") : we have zero interest to ruin other users experience. About the "this is my lawn" thing : As xii5ku said : this is *not* anyone's lawn. We are all invited to play on it. We would prefer to see you joining us to play on the common lawn, but we won't force you : after all, this is just a game, not a war declaration. If you think the lawn gets too crowded and you consider there is not enough room for *you*, just wait for the end of the game, or go play elsewhere : there is a lot of Boinc projects hoping for your precious CPU power. As you seem pretty confident we are all selfish people, you may only think we'd like our "stupid" competition not to be slowed down by a weakly driven project. Truth is : we certainly do not want to do any harm to the project, we absolutely do not want to annoy you. Still... Fact is the server is overcharged. Work units downloads and uploads are slowed down. This *is* a poor experience for us all. You may rejoice : our "competition stupidity" will come to its end in 4 days. Things will then be back to "normal". The lawn will be all yours again. Have fun ! We all are engaged in distributed computing experience, involving some computer skills and... "passion" ? I'm sure you have already tried to give some machine power to strange new Boinc projects, right ? Some of them, in alpha mode, throwed you results in error after hours of calculation... Or disappeared after some days of testing phase... Wasting you precious CPU time and electricity bills. If so : remember you decided to engage CPU/GPU power in a *community* effort for free. You took risks to help foreign searchers to prove obscure conjectures you can't understand without a physics PhD. As you're a computer and Boinc enthusiast, you've accepted your efforts could lead to *nothing", because this is the way science works. If you want to be absolutely sure you won't "loose" any CPU time, do as we all do : sign in multiple projects. You'll be sure to have your CPUs/GPUs always fed, avoiding single-server timeouts and "competition stupidity". (Unfortunately, I can't guaranty you won't encounter useless/bad driven/crazy projects). I'm pretty sure you won't get them right, but please receive my apologies for these hard times you are facing. |
|
3)
Message boards :
Number crunching :
Project has no tasks available
(Message 5402)
Posted 12 May 2022 by [AF>Libristes] ElGuillermo Post:
I agree : there's always some *jerks*, unfortunately. But I'm pretty sure you'll agree they are absolutely not representatives of the Boinc community. Though, I understand the frustration of "normal" Boinc users who actually face server downtimes. This is quite annoying and, believe me, absolutely not intended, as it kills the fun even for competitors. The Boinc servers are quite efficient, but the "average guy" CPU power has increased quite impressively these last years : as the Boinc projects owners can't afford thousand dollars machines/network throughput, competitions are becoming quite a challenge to face :/ I hope the lessons will be learned and solutions will be found to avoid frustrating people :) |
|
4)
Message boards :
Number crunching :
Project has no tasks available
(Message 5401)
Posted 12 May 2022 by [AF>Libristes] ElGuillermo Post: Hi ! "Bunkering" is used in a tactical way for competitions : some teams like to gather Boinc users during a limited time period, asking them to "race" on specific projects. There is no *real* prize to win, this is mainly a fun way to use Boinc and how to optimize your machines/clients to get the most out of them. As the winners are sorted by the credits they gain in a limited period of time, you can try to maximize your score by "bunkering" results. Here is how it works : Let's say there will be a competition between the 1st and the 15th of june. 1. You first have to learn how the choosen project behaves on your machines : is Linux faster over Windows ? Does it help to use multithreading ?, etc. Then you'll configure your machines/Boinc clients accordingly, spending (a lot of) time on forums and testing things to get maximum efficiency. 2. As you want to get the maximum possible credits in a limited period of time, you can "bunker" them. The credits are given on work units validation by the server, so you can ask your machine to get work units *before* competition starts, calculate them all, then hold the results to be sent back to the server until competition starts and get the credits validated during competition time. 3. You have observed the server is configured to wait for a result for 7 days after the work unit has been sent to the machine. On the 24th of may, you ask your machines to get as many work units as they can. Depending on projects and machines, you may get thousands of workunits to calculate. Then you'll prevent your Boinc client to use network : the results are "bunkered". 4. On the 1st of june, you allow your Boinc client to use network again : all the "bunkered" results are sent back to the server. You get a major kickstart in the credit race ! Then you let your machines calculate as usual, sending the results as soon as they are available, till the end of the competition. ... This may sound a little childish :) Let's say it's like sports : absolutely non-essential, but fun to play and watch :) Bonus : these competition times help Boinc users to know and talk to others fellow competitors. We all are involved in Boinc because we like fiddling our computers and give some computing power to science projects... Since I've started to crunch on Seti, more than 20 years ago, I met cool people sharing the same passion from almost every part of the planet :) |
|
5)
Message boards :
Number crunching :
Project has no tasks available
(Message 5371)
Posted 9 May 2022 by [AF>Libristes] ElGuillermo Post: Hi ! Retention (figure 1) is a real thing, but the work units must be returned "in time", so this may sound strange to non-competitive users, but doesn't hurt the project, nor the server in any way (beside the fact that a large number of WUs returned in a short time *could* rise the i/o charge on it). But what you describe in figure 2 is nonsense : what could be the benefit ? Doing such thing, you'll blockade every user that waits for a validation, including your own teammates, as there is no way to "choose" the WUs you're getting. I invite you to talk about the Penthatlon (and the other Boinc competitions) with any participant, from any team : you'll soon discover we all are getting fun crunching on various projects, but generally with the permission of the projects owners/developpers, because, as we like to give some CPU/GPU power to science, we do not want to do any harm to the projects. You'll also see the major projects owners are generally happy to see their apps selected for "runs", getting major power boost to their researches. I've been crunching on Boinc for years now, participating in some competitions. I've never seen people trying to "play" with unfair or destructive tactics, such as "holding the tasks until timeout" : we all first want to help projects to get more results. The only "bad" thing you'll may face is a machine crash with hundreds of WUs still on it... That happens. Fortunately, not often. Obviously not on purpose. Have fun. |
(English)
(Polski)