Message boards :
Number crunching :
SETI Farm deployment
Message board moderation
Author | Message |
---|---|
speedyj Send message Joined: 28 Apr 01 Posts: 8 Credit: 19,947,914 RAC: 35 |
I have a number (340+) of windows machines that I'd like to use for crunching. Does anyone have a quick and dirty method of deployment from the /a installer option? I'm more of a Cisco network guy and not an MCSE by any means. Thanks for any any input. |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
Whatever you finally plan to try, just don't do it on this or any other weekend. You might crash the entire project! May this Farce be with You |
Keck_Komputers Send message Joined: 4 Jul 99 Posts: 1575 Credit: 4,152,111 RAC: 1 |
You should be able to set the installer to install to all of them and automatically attach. It does require a package from MS to modify the installer. I'm sorry I do not have more details. BOINC WIKI BOINCing since 2002/12/8 |
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
I have a number (340+) of windows machines that I'd like to use for crunching. Does anyone have a quick and dirty method of deployment from the /a installer option? I'm more of a Cisco network guy and not an MCSE by any means. Thanks for any any input. Hi, there is some remote rollout stuff at these sites: http://www.shadyshack.dyndns.org/boinc.html http://www.teamocuk.com/downloads/addons/br/ http://forums.overclockers.co.uk/showthread.php?t=17347384&highlight=BOINC+script Hope it helps and good luck. |
speedyj Send message Joined: 28 Apr 01 Posts: 8 Credit: 19,947,914 RAC: 35 |
Thanks for the input folks. I'll give a few machines a shot this weekend and see what happens. |
uba36 Send message Joined: 17 Jul 02 Posts: 74 Credit: 1,159,280 RAC: 0 |
Whatever you finally plan to try, just don't do it on this or any other weekend. You might crash the entire project! Oh no, he did it on weekend ! See status page! |
bobb2 Send message Joined: 5 Feb 00 Posts: 53 Credit: 380,595 RAC: 0 |
This might be appropriate question on this thread. There are some users who are running scripts to create very many hosts on at least two of the boinc projects including seti. This wastes resources and will impact processing in the future. 2 top offenders have 360K and 78K hosts with 0 credits. Ignoring the reasons why some would do something like this, what do you think is a reasonable limit to the number of hosts that a user should be allowed to register? We know that some of you have very large farms out there. Bob B |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
This might be appropriate question on this thread. Well, I would say that 100 is a reasonable limit ... maybe think of something like, if I have 100 productive hosts, you add on anther 25 ... Make is some sort of sliding scale and it should serve those with few or many ... heck, you could even start with 25 as a new initial upper limit ... that would also stop the run away spawning (I think) ... if I have 200 and only 2 are doing work ... well ... |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
Let me take this opportunity to plug one of my favorite themes: lets make this system smarter, in this case, by not applying ad hoc quotas and thresholds, etc. (where the emphasis is on ad hoc numbers like 100 -instead of 99 or 101) but instead rank each client (where the system is enabled to learn about the client base). For example, keep a simple table of the probability that the client will return the next wu with a valid result within N hours and then issue this wu based on this probability. As wu's become available issue them to the requesters with the highest performing machines. Wu's with higher priority, such as those required to complete a validation cuple, can be assigned first, which in turn should reduce the size of the validation-pending tables. Of course nothing is simple, and lots of refinements could be envisioned, but you get the idea. I'm willing to bet the entire project will run faster and we will find ET sooner if this type of algorithm was implemented. May this Farce be with You |
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
This might be appropriate question on this thread. All I can say is there are a couple of 'organization' out there, namely 'Ministry of Serendipity' and 'Sneezy' who turn in over 6,000 classic WUs a day. I'm not sure exactly how that translates to hosts, but lets guess at over a thousand each. IMO, for sake of science, we don't need to alienate (no pun intended) these guys. |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 |
This might be appropriate question on this thread. Assuming that they are not cheats, of course we do not want to aleniate them. However, some of the highest returning users were cheating in one way or another. BOINC WIKI |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
If I have 100 hosts, of which none are returning work ... why do I need to have more? This is back to the "bad" hosts and ill behaved people consuming resources ... if my 100 hosts are all returning work, well tomorrow, let me add another 100 ... if they all return work ... and so on. If the account is just spawning hosts, there is no reason the system can't "throttle" them down. Same thing we are doing with the work units... you pull too many and return junk ... lower the limit ... [edit] even if they have 1,000 in classic, it is going to be a little difficult for them to stand up 1,000 hosts overnight ... |
KWSN - MajorKong Send message Joined: 5 Jan 00 Posts: 2892 Credit: 1,499,890 RAC: 0 |
Something like this is already in the code. This line Average turnaround time 3.02 days in one of your computer listings http://setiathome.berkeley.edu/show_host_detail.php?hostid=852707 Computers with low Average Turnaround times (if I remember correctly, less than 1.00 days) are given preference on result re-issue. I am sure the devs have other ideas along these lines. https://youtu.be/iY57ErBkFFE #Texit Don't blame me, I voted for Johnson(L) in 2016. Truth is dangerous... especially when it challenges those in power. |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
Ok, that's good. A little bit of smarts there. The client you mentioned is the laptop I'm using right now. It completes a wu in about 5000 sec, give or take. It has an average turn around of 3d because I want to squeeze the system for all the credits I can (vanity), and coming late with a low score helps. This raises another concept. Perhaps the credits granted to any one client should be discounted for how late they are, not just zero'd after 14 days. This provides incentive to return results quickly, which should help the next step in the scientific analysis, right? May this Farce be with You |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
This provides incentive to return results quickly, which should help the next step in the scientific analysis, right? Seti WU's are created at Arecibo. There is NO preference that one wu be done before another. For example, they just grab a tape from the box without regard as to the date it was recorded. That's why some WUs are from 2004 and some from 2005, and even some from 2003. Folding at home, however, builds the next WUs from the previous runs(in most cases), so time is more important. |
PhonAcq Send message Joined: 14 Apr 01 Posts: 1656 Credit: 30,658,217 RAC: 1 |
Yes, I understand (I think) what they do regarding tapes. But it leaves me rather deflated. It seems like this is just a random search with no report outs. Kinda aimlessly computing stuff, without understanding it. It wouldn't take so much additional effort to take a set of tapes for the month of Dec, 2004, say, and focus on analyzing them at some level of discretion. Call that milestone 1. Then report what has been learned for each such milestone as we go through time. Then time becomes important and 'we' get more excited about the project. Not to cast dispersions, but late at night, this project seems to be poorly managed, at least as far as the science goes. May this Farce be with You |
speedyj Send message Joined: 28 Apr 01 Posts: 8 Credit: 19,947,914 RAC: 35 |
So if I end up scaling this to 1100 hosts (which I have access to total) and say.....100 of them don't, for one reason or another (network outage in some part of the global network - highly unlikely) return work units, am I improperly presenting myself or somehow throttling my stats despite the actual number of working machines? |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
Yes, I understand (I think) what they do regarding tapes. But it leaves me rather deflated. I don't know why. Each WU (regardless of date recorded) has the same chance of finding ET. Each tape is split into frequency bands, the divided up into WUs that OVERLAP each one next to it so that NO gaussians are missed. |
FalconFly Send message Joined: 5 Oct 99 Posts: 394 Credit: 18,053,892 RAC: 0 |
I'm a friend of KISS (Keep It Simple, Stupid) solutions : Let's just auto-delete Hosts, IF a) it is older than X days (e.g. 28) AND b) it never had any Results validated in that time (= has no Credits attached to it) That way, people could spawn as many Hosts as they want; after X days, all those would automatically vanish to DEV:NUL where they belong. If also wouldn't hurt anyone (e.g. someone installs BOINC, then postpones his upcoming DC career by a month). The machine(s) will simply aquire another Host ID next (first) time they connect, and that's it; this User probably won't even notice... Of course, such a Script could be run entirely manually, whenever the Admins would deem it neccessary to clean up a bit (or feel like pressing a big red button after a tough day ;) ) |
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
This might be appropriate question on this thread. AFAIK, Ministry of Serendipity have been the largest SETI classic producer for several years, although they are most definately very anonymous (Mr One and Mr Two), the word is they were checked-out by Berkeley long ago and accepted as legit. Rumor is that it's something like a backup data-centre for a bank. Sneezy came on the scene a couple of years back and now out-produces MoS. Both incidentally are members of the 'Finest Team in the World' - OcUK ;) |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.