Back at the beginning of August I decided that after having completed 2+ million WU's that I would add 6 other interesting projects to my main Workstation as it is the most powerful in my business. I was laboring under the concept that you could set the Resource Shares in such as way as to favor whichever project you want to get more time. So I originally set up everything so that SETI got 65% of total power of 75% of the 8 CPU's in my primary workstation the balance split between the other projects. But was happened apparently as time as gone by is that the Einstein Project has taken me from 4,100 avg SETI credit level down to 400 while it has gone to over 5,000.

About a month ago I posed the question did the Resource Share actually function as it said it did and advice came through to set everything back to 100 and let it settle out. Briefly the Einstein numbers came down by 10% and the SETI ones just continued their downward plunge to the current 400.

Short of just suspending Einstein is there a way to get the Resource Share function to actually work? 11 years of doing SETI and it only took 2 months to drive down my avg to practically nothing. I used to have 200 SETI WU's in my queue. In the last month I even stopped getting astropulse WU's and very few 6.04's have come up.

Help please?

James

---

Joined SETI@Home in 2001
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.

Short of just suspending Einstein is there a way to get the Resource Share function to actually work? 11 years of doing SETI and it only took 2 months to drive down my avg to practically nothing.

You don't need to suspend Einstein, just set No New Tasks, let your Einsten Wu's complete, as you complete the Einstein Wu's you'll get more and more Seti Wu's, once you're out of Einstein Wu's and your cache is full of Seti, then unset NNT at Einstein, Boinc should get a few Wu's, now set NNT again until you want some more,

The problem is that Boinc will do a request to Seti, and doesn't get any work, if it does a couple of requests it will, but in the meantime it's run through all the other projects asking for work (and getting it) because it didn't get any from Seti on it's first try, eventually Boinc will stop asking for work from those projects,

Claggy

Well, I didn't get BOINC to run two projects 50:50 on a dual core CPU properly, so you probably should not expect too much with more projects and more "complicated" resource shares. BOINC's way of caching WUs from projects in case of occasional "no tasks available" (or even without it) is more than weird.

Basically I don't see any way to get that to work when the cache size predictions doesn't seem to work (see my work_fetch debug log in the last post from my thread).

---

I've been unable to get BOINC 7 to pay any attention to resource share settings, so I get a batch of tasks from my "B" project and suspend them. I release enough each day to match my resource share desire. That blocks work fetch for that project, so I have to release them all when I need more work from the "B" project. That's some intervention, so it's up to you. I use Boinc Tasks, so it's just a click or two rather than the chore it is in BOINC.

The new option for cc_config.xml had some effect with Einstein when I set the half life to 1 day vs. the default of 10 days. It had no effect with WCG.

<rec_half_life_days>X</rec_half_life_days>
A project's scheduling priority is determined by its estimated credit in the last X days. Default is 10; set it larger if you run long high-priority jobs. New in 6.13.1

---

Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.

I'm having the same kind of problem. The only way you can make sure SETI will get its share is to make sure there is enough SETI work in your cache. The only way I can get this working with the BOINC version I have here (7.0.28) is to manually intervene by setting all other projects to 'no new tasks' and than increasing the minimal work buffer. It usually takes a couple of hours before I get enough SETI tasks (because of 'no work available' or download problems) and than I follow the reversed procedure (decrease minimal work buffer and allow new tasks for other projects). It's a pain, but the only way I can get SETI to reach its resource share while also running several other projects.
BOINC does not seem to handle projects that do not always return work at the first request very well.

Tom

---

On my 3 hosts, 2 Quads and 1 i7-2600, I run several projects.
The 2 C2Q each have an NVidia (FERMI) GPU. (GTX470 & GTX480){WIN XP x86 & x64}
The i7-2600 CPU has 2 AMD/ATI HD5870 GPUs. (Cross-Fire DISabled)(WIN 7; 64bit)

SETI has 40% resources, all use Home, as basic setting.

It can take several weeks, even months, depending on availabillity of
work, also the Kind of work done by CPU only, CPU+GPU or GPU-only*
*CPU is always needed to 'feed' the GPU(s).

F.i. Collatz C., Milkyway, Primegrid, GPUgrid, have GPU only work
which isn't easy to compaire to CPU+GPU like SETI/Einstein, etc.

You can use a a program like BOINC-Tasks or similar.
Doing this without a third party program, takes some time and knowledge
of your CPUs and GPUs capabillities. (Mobo; DRAM speed and amount and PSU are
equally important.

The use of (LUNATICs) Optimized applications, increases your
overall efficiency, also has an effect in your resource-share.

Best is to start with a resource-share, you want and just let BOINC run.
B.t.w. I use BOINC 7.0.28, x86 & x64 on all 3 rigs.

You'll soon find out, GPU only projects appear to gain from this setting and
give a much higher RAC, compaired to a CPU+GPU project and a CPU only gets
the lowest RAC or throughput.

Although the FLOPS reading provided by BOINC isn't very accurate, but does
give a usefull 'compairison tool'.

Fred Efmers tool to establish how to run your GPU as effective as possible,
whithout making errors or getting too hot, is very usefull helping to find
the best setting.

It even gets a bit more complecated, as Einstein uses another 'tool' to use
1; 0.5; 0.33 or less(0.25) GPU. Atleast not through an app_info.xml
file. That way Einstein uses 0.5CPU+x.xxGPU.

In SETI, an app_info.xml file is used, not sure how GPU-only projects
manages this.

To make a long story short(er), by trial (no errors) you can change your
resource setting, till you've got the most ideal setting.

I know, with a lot of hosts, this can get you a days job ;-).

Finding out which host is most suitable for 'a project', can help finding
a similar project as 'back-up' project.

Probably not much of a help, but that's the bottom line if you don't want
to use a 3rd program to manage BOINC.

---

Insightful thread, I'm new to all of this and trying to manage a 50/50 share between two projects and for some reason SETI has recently begun to fall out of BOINC favor rather quickly. I value the info being shared here. Thank you!

The reality is the system does work. But many things happen that screw with it. The time period it averages over is more like a couple of months. We humans are an impatient lot and expect it to average over less than a day.

The biggest issue is with Boinc 7.X and multiple projects is when it makes a request of a project it will completely fill the cache from that project. So if for some reason the first project it asks for work doesn't have any, it goes to the next project and fills the cache, it may be some days before it asks for work from that project again as the cache must empty first. So if you have a large cache this makes it much worse.

---

Well based on the comments since my last post 5 days ago I am trying the following and I will report back on it after it has about 2 weeks to generate some results. Previously I had all projects at contacting home every 2 days and to get at least 10 days worth of work. I also set the size of available disk space to 10 gigs. And finally the resource share was all of them at 100. I'm not worried about CPU and or memory setting changes so those are staying at what they were.

So here's the change. 1) All projects get disk space reduced to 4 gigs except for Einstein which goes to 3 as it seems to need lots of space (this an attempt to curb its activity). 2) The WU's cache settings SETI remains at 2 & 10 but all the rest get changed to 2 & 4 with the exception of Einstein with goes to 1 & 3. Finally 3) SETI resource share goes to 150, Eistein goes to 25 and the rest remain at 100.

And at least temporarily Einstein is suspended from getting new WU's while its cache works down.

Hopefully this will result in a level of activity more in line in what I wanted which was to keep SETI over 2000 but allow the others to get reasonable numbers not my current situation which was SETI falling from 4400 to 440 and Einstein briefly at 5000 and the rest all between 200 and 900..


James


ps Does anyone know of a way to have BOINC use its operating data directory on another drive letter? I would be better for me to be able to move the data to an independent drive from the boot/application drive. I mention this because I have already gained benefits from operating BoincLogX data directories on another drive as allowed them more space and to run faster.[/i]

---

Joined SETI@Home in 2001
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.

ps Does anyone know of a way to have BOINC use its operating data directory on another drive letter? I would be better for me to be able to move the data to an independent drive from the boot/application drive. I mention this because I have already gained benefits from operating BoincLogX data directories on another drive as allowed them more space and to run faster.

Yes. It's easy and safe to do (with care), but you may find the instructions a little alarming.

1) Make sure you have two things to hand:
a) A note of where your BOINC data currently is.
b) A copy of the BOINC installer (setup program) for your preferred version.
(for the location, see the first few lines of your message/event log when BOINC starts up)

2) Shut down BOINC completely, including running science applications.

3) Uninstall BOINC - that will remove the BOINC program itself, but leave your data and science applications untouched.

4) Move your entire BOINC data folder - the one you noted at (1a) - to the new location. Move all files and subfolders.

5) Run the BOINC installer. On the setup options screen, click the 'Advanced' button. From there, you can browse to your preferred location for - separately - both programs and data. The important one, of course, is the data folder. Make sure you tell the installer the location you moved it to at (4).

6) Start BOINC again. It should pick up where it left off, as if nothing had happened.

(please don't listen to anyone who suggests that you install BOINC first, and move the data afterwards)

So here's the change.
1) All projects get disk space reduced to 4 gigs except for Einstein which goes to 3 as it seems to need lots of space (this an attempt to curb its activity).
2) The WU's cache settings SETI remains at 2 & 10 but all the rest get changed to 2 & 4 with the exception of Einstein with goes to 1 & 3.

And how do you do that? (since these separate settings are not possible)
- "WU's cache settings" is total work (sum) from all the projects (in calendar days)
- The "disk space allowed" (3 values, most restrictive is used) is total disk space BOINC can use for all the projects in this computer

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

So here's the change.
1) All projects get disk space reduced to 4 gigs except for Einstein which goes to 3 as it seems to need lots of space (this an attempt to curb its activity).
2) The WU's cache settings SETI remains at 2 & 10 but all the rest get changed to 2 & 4 with the exception of Einstein with goes to 1 & 3.

And how do you do that? (since these separate settings are not possible)
- "WU's cache settings" is total work (sum) from all the projects (in calendar days)
- The "disk space allowed" (3 values, most restrictive is used) is total disk space BOINC can use for all the projects in this computer

What???? Ok now I'm floored. I assumed that the each projects settings were independent of the other projects settings especially as some projects had settng for things not present in other projects. If all the projects use the lowest setting possible I now have to go through and reset all the project settings to what they were before I started experimenting trying to get the BOINC program to share the resources equally because after I added the extra projects the Einstein Project was stealing all the time and the SETI program went to the scrap heap in terms of WU's completed. Even now I am just starting to get back to 1000 avg on SETI but I haven't gotten an Einstein in 10+ days.

---

Joined SETI@Home in 2001
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.

I changed all the settings to be equal again and still no Einstein WU's coming through.

My SETI's avg is coming slowly slowly back up its about 900 when it was at 4400 on Aug 5 with I added the other projects.

Someone at Einstein suggested that I suspend getting new SETI work the way I had suspended Einstein work to get SETI work again because I had dropped off to just a few SETI WU's coming through which is how I went from 4400 SETI avg to 300 in 6 weeks and in the last 4 I've gone back up from 300 to 900 because my Einstein has gone from 6000 down to 1500.

But I have noticed that I was getting a bunch of the astropulse long running WU's in July and for part of August but I havent gotten 1 since the end of August and only a few 6.04 have come through since I started getting SETI's again. Is there a reason why I wouldnt get get 6.04's or astropulses anymore?

And am I going to screw myself up again if I temporarily suspend getting SETI WU's for some Einstein ones will start coming. I dont understand why the BOINC system just doesnt do a TSO and do so many for each project I've got enough computer power to do it.

James the puzzled.

ps I really though the fact that BOINC maintained a separate xml for each projects settings plus the individual default, home, work, school because I was using those settings to control how each machine running BOINC was using its resources.

---

Joined SETI@Home in 2001
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.