The Gridcoin mining experiment - host normalization continued

Time flies, it has been two week since the last update so let's have a look at the data evolution.

Purpose of the experiment

I continue this experiment because it provides interesting facts. The ones who are very involved in gridcoin know that RAC can change and is not predictable for various reasons. This experiment shows this with real life data and can help newcomers by example to see how complex gridcoin mining is and give experience users some data to look at and confirm their understanding and insights.

This data is gathered on a virtualized Microsoft Windows Server 2012 R2 Standard x64 Edition running on a Intel(R) Xeon(R) CPU E5-2630 0 @ 2.30GHz [Family 6 Model 45 Stepping 7] core with 4GB of RAM. The computing preferences are set to use at most 50% of the cores at 100% of the time. The project being crunched is VGTU.

Data gathering

As I've been busy with this experiment for such a long time I tend to forget to explain about how it all got to this point and only explain the differences since the last update. One of the biggest reasons I started this experiment is because I knew from prior BOINCing experience, before gridcoin, that RAC is not something that goes to regime as it is often oversimplified. The thing is that there is no data available to show this. If you're only BOINCing this doesn't really matter because you're practically donating computing resources. As it currently stands gridcoin mining is built around RAC, it is a central pillar of it.
BOINC projects have quite some freedom to change how they present and calculate data but most keep defaults as far as RAC goes. The host page for this experiment's host can be found here. As you can see only the current RAC is available from the host page. There is no way to see the history of it. The tasks themselves have some history but that is also limited somehow. From the task overview you can navigate to next 20, and do that a couple of times but then the history also stops.
The biggest work for this experiment is the gathering of the data. I manually copy paste all of it in a (somewhat) timely manner to be able to make up the history and the charts to be able to see patterns, show how things evolve and help the community by providing data to look at and think about.

Data loss

As mentioned in the last update there has been some host downtime. Now that the data gathering has continued the effect of this is better visible. The green series, points per seconds, has a clear drop. The RAC however has more smoothness in it. The exponential drop shows itself as almost straight down, but the build up, which started not from the bottom but from a certain level, is better visible. You can almost see the continuation.

Host normalization

The host downtime came a bit at an unfortunate moment. Nonetheless we can see that when the team RAC bottomed the RAC was near its top. It will take some further data gathering to see if the correlation continues. It seems that the RAC is going down again while the team RAC is going up.

Top 10 users

This is an overview of the top 10 contributors on VGTU. What's interesting here is that the pools keep increasing their RAC so the crash is not really affecting the crunching efforts, which, from a scientific point of view is superb.

Here it is also clearly visible how stopping with crunching doesn't cause the RAC to drop to zero immediately. @dutch's RAC is showing a nice exponential decline.

This post has most info about the first chart: The RAC experiment progress and request for help

This post has the motivation for the second chart: The RAC experiment continues

---

Posted on Utopian.io - Rewarding Open Source Contributors

---