Shifting Gears - Moving away from F@H, need suggestions

M

Mtnduey

[H]ard|DCer of the Month - Nov. 2013/Nov. 2014
Joined
Mar 24, 2003
Messages
1,205
Well, I had hoped to make a run for taking down Anonymous for #1 in F@H. Then PG had to go and make a bigger mess of an already imperfect system leaving me with less than half the point generation I had before. I don't think I need to go into specifics, but if you look at my folding history it pretty much speaks for itself as to why I might be a shade disgusted with PG.

MY FOLDING STATS

So it's time to find a new project(s) and I'd like to find a spot on the [H]ardest team out there. Any suggestions on what to contribute to next?

My farm is entirely CPU based, 2P only, mostly E5-2660 but a few E5649's in the mix. I'm running RHEL 6.x, headless (ssh) text console only (none of that gui $@%#).

I'd like to keep it single project but I can split it up a little if needed.

Project must allow me to configure a proxy server from the application config as my systems do not have internet access on their own.

Thanks for the help :)
 
If you are looking at BOINC (and RHEL may be tough and require the Linux experts of the team), WCG is probably the closest home for FAH users or potentially GPUGrid. GPUGrid has work for pretty much all hardware, but availability may fluctuate and the CPU work doesn't score well. WCG is a low scoring BOINC project, so if you compare its points with all BOINC projects, it can be a bit disgruntling. However, it has one of the best reps in the BOINC world, a great staff, reliability, does only humanitarian work, and tends to have the largest portion of our team. Compared to FAH, our team is small potatoes. When you have corporations the size of SONY and IBM with teams, it can get pretty challenging. So far this year our team has done ~31 million BOINC points (WCG points would be 7 times that if you are converting). IBM's team has done 1.18 billion BOINC points already this year. Considering that WCG is only CPU (including ARM) right now, that is a lot of computing.

WCG is also a great gateway project to wet your whistle for other projects... :D

You will also find that since the majority of BOINC projects are one work unit per thread/core, RAM tends to be in more demand and typical multi socket rigs designed for FAH tend to be lacking. Keep an eye on that.

You can sign up for WCG here: http://www.worldcommunitygrid.org/reg/viewRegister.do?recruiterId=338542
Then join our team here: http://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=BP5XNJBR9N1

And no matter what project you choose (if BOINC based) make sure to attach to WUProp. It is non-CPU intensive and won't get in the way of other work. It will reside in memory pretty much the whole time. It collects data about hardware and projects which you can actually use the results from their site to assist on project planning.
 
Last edited:
What about the Folding + mining project? or is that shafting people on points also..
 
Folding + mining? Are you meaning the FoldingCoin thing? If so, that is just FAH still.

If you mean BitcoinUtopia, that isn't doing any direct science and they are a for profit outfit. CPU mining would be terrible idea and so the contributions to that would be a complete waste of resources.

There is still CureCoin as well, but that is like FoldingCoin where you are still just folding at FAH. Both of those folding gimmicks are not really panning out with value and FoldingCoin would still require you to switch teams.

In BOINC there is still one of those gimmicks too called Gridcoin, but yes they require you to be on their team.

If you had ASICs to run... you could do unbelievable amount of BOINC points at BitcoinUtopia. It is just crazy the points they award those. Most long term crunchers try to ignore those points when comparing across projects....
 
Mtnduey, welcome. :cool:

BTW, I'm Leonardo at Folding@Home, rather was at F@H (#20 overall).

I was there for nearly 15 years. I saw the light and joined the [H]ardOCP crunchers. If you want to see all cores in your AMD 4Ps at 100%, I recommend Rosetta and World Community Grid.

Join the team, crunch away, and have fun. I am. :)

Update: I reread your starting post. Regardless, both Rosetta and WCG will fully light up all cores on your 2P or and any other CPU optimized machines you throw at them.
 
Last edited:
Rosetta is another good solid project with a good admin team. Just be careful monitoring your RAM as each work unit will want up to 512MB.
 
Hi Mtduey

One thing I wish to point out about Boinc there is a bit of a learning curve to Boinc, I would start out with one of the more dependable projects to start with and get them all going on it and start to get comfortable with it. The projects I would recommended for a person that has as many rigs as what you have and me not knowing how much ram you are running per core / thread are.
#1 World Comunity Grid - Humanity / medical
#2 Poem@Home - Medical
#3 Primgrid - Prime # search
These 3 ^^^^ are the easiest most dependable to start with
if you have plenty of ram Roseta@home - is also dependable but can be very resource draining if it is not set up properly just as Linden I recommended it to him without thinking about the ram requirements for it and he had a bit of trouble to start. But with help from the others here he got it running well, I have seen it use up to 60 GB of ram on a 64 core rig before so it can and will use allot of ram.

Once you get some of the more dependable projects up and running start playing around with Boinc there are team challenges and project challenges that can be fun to play with once you get used to Boinc. I would recommend using Bam http://boincstats.com/en/bam/ to start with because you can easily control all of of your rigs from any one rig with it but once again there is a learning curve that goes along with it.

As you well know I run Linux but I am a Linux dummy but fortunately brilong has a few tricks up his sleeve that may be able to help you out as you transition over to Boinc I do not know if Linden has learned any tricks yet or not but as you can see I am trying to pass the buck here. :D
 
Thanks, Grandpa, for bringing up the learning and RAM demands. You are definitely right.
 
I forgot to mention RAM.

Every rig I have has at least 96gb of RAM, most are well over that mark.

Grandpa, as you noted I do have a fair number of machines. With that automated operation once configured is crucial as I really don't want to babysit my machines.

Does this change your suggestions?
 
Well it means that Roseta should be no problem for you :) and as far as Bam goes you can controll all your rigs from any given computer with an internet connection you just need to be able to log into your account from it. There is a learning curve to it but hell I learned it and I am not the best at learning things. :D

As you can see below I am controlling 10 rigs using Bam and I can do it from anywhere. You can connect or disconnect, suspend ,remove etc to any project. It is a bit more complicated than that but once you get it figured out it is simple. :cool:

Bam_zpsqrsagg0x.png
[/URL][/IMG]
 
For my rigs the internet connection is pretty restrictive (hence the need for proxy configuration).

Through the proxy I can have the machine download and upload work units no problem, but the proxy is how they reach the internet.

The machines are on a restricted VLAN behind multiple layers of firewalls, intrusion detection systems, etc...
 
The BOINC client supports proxy, so as long as BOINC is setup properly, the projects shouldn't be an issue. BAM is controlled from a website and not a program, so whatever computer you want to control it from just needs to be able to log into a webpage. I hope that helps clear the proxy thing up. I'm not a proxy expert, so unless I missed something, this should be a no problem issue.

Now, if you were to use the BOINCtasks program to monitor things, then that may be something to consider. RFGuy_KCCO has a thread on that app and is basically HFM.
 
Mtnduey said:
Well, I had hoped to make a run for taking down Anonymous for #1 in F@H. Then PG had to go and make a bigger mess of an already imperfect system leaving me with less than half the point generation I had before. I don't think I need to go into specifics, but if you look at my folding history it pretty much speaks for itself as to why I might be a shade disgusted with PG.

MY FOLDING STATS
Click to expand...

Ouch! Under 4M PPD and you're giving it all she's got? :D

I'll take Grandpa's passing of the buck. I imagine you might already have a trusted host in your cluster; if not, I would set up an SSH trust from one node to the rest as this will greatly ease administration. Like you I prefer not to touch my nodes very often.

The most recent thing I did was disable BOINC on 8 nodes and install the distributed.net client (dnetc) so I could run the OGR-28 project. We are kicking it [H]ard over there. Since NickOfTime and I have joined, we've gained 51 positions at 28.99 points per. Each DC-Vault project has a max of 10K points and we're at 8,782 currently in 43rd place. If you've not read up on DC-Vault, I would recommend it.

I would recommend BOINC 7.2.x or higher for your clients since I was having issues with the older 7.0.x clients on some of the newer projects. I also use BOINCTasks on a Windoze VM to manage my clients when I cannot accomplish things via SSH (aborting thousands of tasks at a time is much easier).

I used to be RHEL-only, but I switched to tear's Ubuntu 12.04-based image a long time ago and I've gotten familiar with both. I've not run BOINC on RHEL, but it shouldn't be too hard. Most of tear's work on the Ubuntu image helped more with F@H.
 
I'm not sure on RHEL 6.x - it might have the client in one of the extra repos. For CentOS 7 I would recommend compiling it. I started a guide on how in the guides section, however I have not yet got a chance to extensively test my result (it seems to be working).
 
EXT64 said:
I'm not sure on RHEL 6.x - it might have the client in one of the extra repos. For CentOS 7 I would recommend compiling it. I started a guide on how in the guides section, however I have not yet got a chance to extensively test my result (it seems to be working).
Click to expand...

You need pictures so old guys have an easier time... :D
 
:D Pictures of me typing in a terminal aren't the most exciting thing.
 
oooh, must use restraint not to completely derail this thread :p
 
brilong said:
Ouch! Under 4M PPD and you're giving it all she's got? :D

I'll take Grandpa's passing of the buck. I imagine you might already have a trusted host in your cluster; if not, I would set up an SSH trust from one node to the rest as this will greatly ease administration. Like you I prefer not to touch my nodes very often.

The most recent thing I did was disable BOINC on 8 nodes and install the distributed.net client (dnetc) so I could run the OGR-28 project. We are kicking it [H]ard over there. Since NickOfTime and I have joined, we've gained 51 positions at 28.99 points per. Each DC-Vault project has a max of 10K points and we're at 8,782 currently in 43rd place. If you've not read up on DC-Vault, I would recommend it.

I would recommend BOINC 7.2.x or higher for your clients since I was having issues with the older 7.0.x clients on some of the newer projects. I also use BOINCTasks on a Windoze VM to manage my clients when I cannot accomplish things via SSH (aborting thousands of tasks at a time is much easier).

I used to be RHEL-only, but I switched to tear's Ubuntu 12.04-based image a long time ago and I've gotten familiar with both. I've not run BOINC on RHEL, but it shouldn't be too hard. Most of tear's work on the Ubuntu image helped more with F@H.
Click to expand...

yeah I added more machines and still couldn't get half of the old PPD. Sad eh?

I do have a trusted host that has full internet access. For F@H I configure a proxy IP & port number for that trusted host to act as the pass-through. I was hoping for something similar in the next project for the sake of simplicity.

I'm pretty set on RHEL 6.x (currently on 6.5) as I already have a lovely standard build that I use, and I've really come to like Tuned.
 
The boinc is just a different animal all together. The projects workload are very different from project to project.

My farm was completely setup for just F@H. Alot of CPU, little RAM, all running on USB flash drives. Some of the projects required more memory than I had whiched cause Disk I/O via swap file. That lead to the OS locking up into unrecoverable read-only mode. I've had to reimage every box I have. I've added memory to most boxes, and killed now 3 of my USB thumb drives. Watched my AMD4P 61xx normally folded at 900W, hitting ~1,100 on some projects, tripping my UPS.

I've had a rough start... You'll want to get BoincTasks, pick a single project until you have management down. I'd love to get a thread started just on the commandline admin for the BOINC service. Currently I got one client just stuck in "Downloading" mode for 24hrs.. nothing seems to kickstart it running.
 
Last edited:
There, I gave you some pictures GilVirPa - I hope you are happy :D
 
I saw the screen shots.. but where are these pics of you at a terminal?
 
Someone needs to tape a pencil to its hoof so it can press one key at a time - that mean Gilthanis not helping it out :D
 
Wow, good luck with jumping on the BOINC wagon Mtnduey, it takes a little time to get used to having one client that is the central of a bunch of different projects.

How many ponies can Gilthanis stuff into one thread ?
 
runs2far said:
Wow, good luck with jumping on the BOINC wagon Mtnduey, it takes a little time to get used to having one client that is the central of a bunch of different projects.

How many ponies can Gilthanis stuff into one thread ?
Click to expand...

As many as it takes.... :D
 
You must log in or register to reply here.
Back
Top