U2 ...... there are other/older posts even i did understand on the subject. U need to go to the BOINC profile and tell it how much you want cached thru the Contact Frequency option: Default BOINC Profile . Find the "Connect to network about every" field. The default is 0.1 days, equal to 2.4 hours. Enter 3.0 there and not more than 7.0 as per our earlier exchange.

PS Your brethren might be in The Lazy Slugs Team

Bouna Notte

I wouldn't normally go beyond 3.0 for this setting. After that, you not only risk missing the deadline on some of your WUs as retsof mentioned, you'll also delay other WUs from getting a timely quorum. Anything between 2.0 and 3.0 will be plenty to deal with normal outages (and unplanned ones) on the WCG servers, your DSL/cable connection, and such. This seems to be more true with the 5.4.9 as I am starting to believe that the time to completion values are much improved in terms of accuracy.

Hi Guys,

First, my thanks to Sekerob for pointing me in the direction of using BOINC (I was using UDAgent). From the point of view of any cruncher that doesn't want to be connected to the internet all the time the ability to cache workunits on the hard-drive is a big plus with BOINC.

Your responses quickly went over my head. However, my summary of the conversation would be: Watch, don't over-do the caching or problems may arise - you have max 7 days to report back crunched units, sometimes an 'unexpected' additional unit will get itself downloaded and throw timings out a bit, other unexpected happenings can do the same too. On that basis I took a middle road on the disagreement and opted for 3.5 days.

As I've now had a chance to test that over a period of time I would say it works fine and I can tinker with it to try and achieve a nicer balance between what units I think I can manage and what units a computer thinks I can manage.

Sekerob did mention a, "Find the "Connect to network about every" field". When I first read the comment I didn't notice the link he had provided and spent some time at my local computer searching BOINC settings for this mysterious 'Connect to network about every' setting.

So that new users don't fall foul of the same misunderstanding I would add to this conversion: That setting exists at at the World Community Grid (WCG) website, there are webpages there where you can make settings that can be used to influence how BOINC (on your local machine) will interact with WCG servers. So, my own advice would be to visit and study those settings a little when you first install BOINC. Doing that will start to make clearer what the 'Activity' menu settings in the 'BOINC Manager (local host)' on your own computer are referring to. To get to the those settings you will need to home in on the 'Device Manager' sections of your WCG account pages.

I would also add that my early experiment with BOINC and caching work-units didn't produce the kind of activity that I expected; even after adjusting settings at WCG. Units were going up and coming down in a manner that I found confusing. I thought I had set up a queue that would complete itself before up- and down-loading of old and new units. It didn't work that way. (I can see now why trying to do it that way would not be the best of ideas.) However, the key to getting a queue (more or less) complete before reporting is to use the 'Network Activity Suspended' setting on the BOINC Manager 'Activity' menu on your own local computer. Enable the 'Suspend' setting and you effectively give yourself manual control over when BOINC will communicate with WCG.

The way I found best to run it, was to let all but the last unit complete. Then, some time during the processing of the last unit in the queue, up- and down-load the completed and new units. This is done in BOINC Manager by clicking on the 'Updates' tab, highlight the 'project' that you want to send results on and then click the 'Update' button. Provided a connection can made completed units then go up and new ones come down. Simple once I got the hang of it.

However, there is a further stage that might be found confusing. Having sent completed units up and got new units down a list appears of items "ready to report". Baffling, haven't I just done that? (The more experienced members here would have to comment on why this happens and what the reasoning is behind it as I don't know myself.) To 'report' those items you need to once again click on the 'Update' button as described above. I would add that in practice it seems best to leave this for a while after having done the initial up- and down-load of completed and new units.

After having done that, the one thing needed to remember is to, once it has completed, upload and 'report' the very last unit of the (old) queue you had before you download all those lovely new units.

Last comment for BOINC beginners (like myself). If I've explained things well enough above for someone to understand the 'Activity' menu selection regarding network activity. Then, I hope, beginners will now see that when that menu refers to 'preferences' the 'preferences' that it is referring to are those that are set by you the user at the WCG site. With that key it shouldn't be any real problem to work out how the 'Activity' (menu) 'Run' selections slot together.

Now my questions for more experienced crunchers:

What is a 'quorum'?
What are 'priority units' and why might I land with one unexpectedly?
What are 'claimed' and 'granted' points?

Thanks again.

The "reported" function could be a function of the project, with behavior as you have noticed. I have run a couple of others that do the upload, and "report" automatically without an extra click being required.

The quorum is the number of workunits that must be averaged to obtain a result. Here, it is 3 normally. I have seen other projects with 1, 2, and even 5 I think. It depends on the scientific accuracy needed. These bio projects are a guess, so some redundancy is needed. The mathematical projects can get by with 1 or 2 since they are numerical calculations.

Claimed and granted can be seen in the detailed boinc device statistics. Suppose you run a workunit and "claim" 30 points for the effort. Somebody else might get 40 and somebody else turns in 50. It all depends on the speed of their computers. All 3 of you would be granted 40 points on the average. (I have estimated the average and do not know the exact algorithm used.) This is an advantage for somebody who always has a point gain each time.

I have seen workunits pop up with a deadline of 1 day. I think these are the priority ones ... probably workunit #3 repeat or #4 that is needed quickly to complete a group. If your computer has sent in some workunits without error, you are more likely to get these. If you have sent in a bunch of errors due to memory problems or excess overclocking, don't expect to get one.

(The BOINC points are later converted to WCG points, but that's something else. There's a similar averaging that takes place on the UD interface. BOINC and UD run separate pools of workunits, and are not mixed.)

If u hit the update button once (in the project tab of BOINCmgr), it will pull the profile update across, after which your BOINC will know permanently what to do.