9th Annual BOINC Pentathlon 2018 Strategy & Preparation

I plan to fill up all my rigs with as many Universe tasks as I can when I am done with it.
 
Don't think I'll be going the VM route, but if someone does you don't need to start that many. Just start one. Get it ready with BOINC installed and the project attached. Save the image. Then load it up with as many tasks as you can. Shut the VM down, delete that instance and open the saved state. Rinse and repeat.

If I wasn't so lost on VMs I would do it on a Windows box. But man, that shit is confusing as hell.
 
I know asteroids@home doesn't really use GPUs better than CPUs, but with my GPUs sitting idle waiting on a project to be announced that GPU I went ahead and started crunching on my idle GPUs on this project. I got a deadline of the 15th so if this project happens to be selected I've got a nice bunker of GPU projects that isn't effecting crunching on the other projects right now.

Food for thought.
 
Maybe you might have forgot to abort them until the 13th... it happens, right?
 
Looks like NumberFields hourly output for most top teams are down. Either bunkering or move to Rosetta and I'm betting it is Rosetta. We have the highest output for the most recent hourly report.
 
Keep in mind that "salting the crops" can also fuck over people, such as myself, who are trying to bring the team up in rank. I can't control what you guys do, let alone other teams, but I don't support that kind of bullshit. I do have some "dead" work units due to accidentally killing off that BoincOS system yesterday and it's really annoying me as I always abort all WU's I'm not going to do. To each their own though.

In other news, since my last update, I've added four more more systems to Universe and am going to try to add another one in a minute if possible. They're all running Linux, since I see to recall that's more efficient for this project. Will also be interesting to see how the 2P-5118 fairs against the 2P-4110 systems in terms of turn around time.

Normally I hide my systems; but in this case I'm leaving them out in the open. https://universeathome.pl/universe/hosts_user.php?userid=6403
 
ChristianVirtual said:
Wonder if we should use the chance getting #2 in Marathon ? Looks in reach
Click to expand...

I hope so. Our team RAC is up to 2,060,000 and we're the 4th highest output. My RAC is up to 792K and theoretically 7th highest output in the project... we'll see if that's accurate when those stats refresh.
 
2P-6126 is now online with Universe. Wish I had more Xeon Skylakes to compare. We should be getting in a 4P-8160 Platinum; but I'm not expecting to have much of a chance to put it through its paces before I start doing proper work things with it. Will try though. As for me... it's time to go home.
 
fastgeek said:
Keep in mind that "salting the crops" can also fuck over people, such as myself, who are trying to bring the team up in rank. I can't control what you guys do, let alone other teams, but I don't support that kind of bullshit. I do have some "dead" work units due to accidentally killing off that BoincOS system yesterday and it's really annoying me as I always abort all WU's I'm not going to do. To each their own though.
Click to expand...


Yeah i was not supporting the idea so much as thinking about the concept in general out loud and how it would play out. Personally i will not be doing this and i don't think it fits in with the spirit of DC I hate to lose any work units and have paused tasked a few times already on machines during this contest to make sure that i finish tasks from other projects (outside the contest) that were nearing deadline. I say we keep it classy and kick ass anyways.
 
Last edited:
As an eBay Associate, HardForum may earn from qualifying purchases.
I already got a GPU bunker of these units. About a days worth so far.

atp1916 toss your GPUs at this now. Go ahead and work on a bunker.
 
If you decide to run both Rosetta and Asteroid tasks on the same rig, make sure you can block the Asteroid tasks from uploading while getting the Rosetta work to upload.
 
Gilthanis said:
I am forecasting that the City Run (3 days) will be Rosetta@home again like it was 2 years ago. I would be shocked if it didn't get the most votes. Technically two GPU projects could get chosen still but I'm betting it will be a "GPU project" and Asteroids@home. Just my predictions. I would love to see Amicable Numbers be chosen given the newfound advantage miner boxes appear to have there.
Click to expand...

Maybe I should buy some lottery tickets....lol
 
Since our GPUs are sitting idle might as well use them.
 
That is what most people are thinking right now. Also, I'm betting most people are only going to use the GPU portion until the other GPU project is announced and then make adjustments as needed. This is where having multiple clients really shines. That or using VM's for the CPU work units and leaving the bare metal host client waiting for GPU projects. Sure, you can manipulate firewalls and such but most people only have a couple rigs to juggle. That and I don't like mucking with my firewall...lol.
 
pututu said:
If you decide to run both Rosetta and Asteroid tasks on the same rig, make sure you can block the Asteroid tasks from uploading while getting the Rosetta work to upload.
Click to expand...

Can you provide hostfile instructions and IP's for both projects. I am already moving my 2 gpu systems over to both but currently just suspending network
 
Anyone tell me how to allow some projects network access while blocking others?

I've got about 200 unverified WU for Universe right now, want to start a little bunker on the others
 
See thread here: https://hardforum.com/threads/2018-marathon-run-numberfields-home-support-thread.1959633/

I Windows, I think you need to add the 127.0.0.1 asteroidsathome.net in the hosts file. I've not tested on rosetta yet.


upload_2018-5-9_12-51-29.png
 
atp1916 said:
I've been looking at C621 setups and am surprised by how comparable the cost would be vs. getting a v3/v4 setup going:

Supermicro MBD-X11DPL-I-O (open-box) - $323
x2 Xeon Silver ES QLH2 8c 2.4ghz - $400
OR
x2 Xeon Gold 6130 ES QL1M 16c 1.8ghz - $600

fastgeek In your experience, are the Skylake-SPs better than comparable core/ghz count v4s?
Click to expand...

To be honest, I haven't paid that close attention to how much Skylake-SP outperforms Broadwell/Haswell-EP. If there are any specific test you'd like to see, let me know and I'll see if I can run them for you. Until then, I've allowed my systems to be shown on Number Fields, so you're welcome to take a look if you'd like. (Mix of Sandy, Ivy, Haswell, Broadwell and Skylake)

https://numberfields.asu.edu/NumberFields/hosts_user.php?userid=67667
 
As an eBay Associate, HardForum may earn from qualifying purchases.
I was in several minds as to what the hell to do next.

I'd like us to not to give up another place in the swimming event so I've decided to stick with Universe until close to the death.

In addition, I've filled my boots with Asteroids work, parking the cpu tasks but crunching the gpu ones. It appears that the boinc limit of 1000 tasks max applies to any combination of cpu/gpu downloads when taken together in total.

The Universe/Asteroids plan effectively means that I'll be doing very little on Rosetta other than maybe a small window between the Swimming and City Run events.

It looks like the Cross Country event is unlikely to be Seti given that the timing overlaps that project's Monday/Tuesday server shenanigans. There again, this may be what they want us to think ....

fastgeek said:
See this post from phoenicis showing which IPs to block for various projects.

https://hardforum.com/threads/9th-a...strategy-preparation.1956723/#post-1043606087
Click to expand...

I've amended the post to correct the Seti error and confirmed that the Asteroids IP is good. Everyone please be cautious with respect to the two IPs for Rosetta as they've not received a recent check.
 
phoenicis, maybe you can create a thread in this sub-forum about IP blocking; then perhaps rgMekanic can see that it's made a sticky. That way we always know where to look for this information in the future. :)
 
phoenicis said:
I'd like us to not to give up another place in the swimming event so I've decided to stick with Universe until close to the death.
Click to expand...
I'll have one rig crunching Universe just to keep SG from catching us up at current run rate. If they have sudden bunker, that will be extremely diffuclt for us to do anything with quorum of 2.
 
phoenicis, prep your Rosetta work in a VM to run on CPU's...let the host run your Asteroids...then it is 1,000 for the bare metal client and 1,000 for the VM.
 
Can confirm the IPs for rosetta do work.

edit
I blocked both IPs on my firewall. Not sure if one or the other or both are needed.
 
fastgeek, we seem to be holding onto the marathon 3rd place spot fairly well. Unless the other teams are bunkerings it, which I doubt we need to focus on something else right now.

I think we should load up some rosetta@home tasks and dump them when the sprint starts, but leave current rigs on Universe@home. Maybe even move some to Universe@home. While we most likely wont claim another position, the team behind us is closing in on us and if we lose that position we'll definitely lose our 3rd overall. Although we will end up losing that position once Team China falls into the #3 spot on Universe which they seem to be working hard at doing relatively easy right now.
 
All of the Linux boxes running Universe aren't on BOINC client, so can't move them from here (home). I think most of the systems running NF are on BC, so I could move them. Right now I seem to be doing 1.5-2M PPD on NF and 1M PPD on Uni.

Of course I just banished Rosetta from some of my systems as it was hogging up an obscene amount of the limited disk space a few of these servers have. (no idea why I did what I did... but too late to fix that now)
 
You must log in or register to reply here.
Back
Top