The Gridcoin mining experiment - VGTU errors and pomegranate
Computation errors
As mentioned in the last update a lot of VGTU tasks were giving errors. I let this continue for a couple of days to clear the backlog but then decided to contact the VGTU project. I quickly got this response from Dr. Vadimas Starikovičius
Hi,
Yes, there were changes to the tasks input.
We are constantly working on the current research problem. Although, You don’t see changes in client applications (versions are the same), generation of work and assimilation/processing of the results are changed from time to time.
We have fixed the bug in work generator on Monday. However, some erroneous work units may still be there…
I don’t know exactly which to cancel them. Don’t worry, no time is lost: application simply fails to start.Best regards,
Dr. Vadimas Starikovičius
In the meantime the backlog has been cleared and no more errors occur.
Because of this change in work generation computation time has risen. Over time this should not influence the points rewarded per time unit but in the short time you can see it does as can be seen in the peak forming over the last two weeks. My explanation for this is that this is again host normalization at work. Several variables are at play: different host take different time to finish the work, different hosts have different amount of resources assigned to BOIN, different hosts have different backlogs, ... In the beginning host normalization gets messed up because the first new tasks have no competition, other hosts are still computing and few results are available. The results that are available are rewarded more points again making them more important in host normalization. The result is that points rewarded goes up. After a while host normalization catches up and pushes the reward back down. Again we can see the pts/s are moving quite linear while RAC is showing exponential shape. We can expect this behavior to be a temporary phenomena that will correct back to regime.
Team RAC
Team RAC is bit harder to correlate here. Obviously the previous described logic has influence here as well, but it must also be noted that crunching power keeps increasing.
Top 10 users
As I was presuming previously pomegranate is back on VGTU. We know this account has a large crunching power so we can expect it to grow considerably. As pomegranate is a sensitive subject to touch I wish to state I'm not going to repeat my opinion about it, I only want to show what the data gathering is bringing. For people who are mining and are crunching BOINC tasks for the money you can expect this account's RAC to go up considerably. This will influence team RAC and thus your magnitude and earnings.
A second observation is that while pomegranate is going up pools are going down. This might however be related to the host normalization. More data is needed to make a correlation here.
About the experiment
If you want more background on the experiment I refer to these posts:
Posted on Utopian.io - Rewarding Open Source Contributors