LLRnet/PRPnet rally January 3rd-10th

[mdettweiler]

Originally Posted by gd_barnes

Thanks to everyone who participated in our 2nd most successful rally ever on a pairs processed basis and most successful in nearly 2 years! We found an outstanding 25 primes for n>890K!

In fact, I wonder if in terms of actual CPU-hours put in, this rally was the 1st most successful. Our January 2009 rally cranked out an enormous number of pairs, but they were very small pairs (<400K IIRC) that tested a lot faster.

Dave, would it by chance be possible to add a "Daily Score - All Time" graph to the Dashboard page? We currently have "Pairs Returned - All Time" but it's a little hard to make a comparison on that basis since the graph is dominated by the occasions when we had large amounts of resources crunching very small tests. A score graph could tell us quite handily whether this was truly the biggest rally in terms of what people brought to it.

[gd_barnes]

Originally Posted by mdettweiler

In fact, I wonder if in terms of actual CPU-hours put in, this rally was the 1st most successful. Our January 2009 rally cranked out an enormous number of pairs, but they were very small pairs (<400K IIRC) that tested a lot faster.

Dave, would it by chance be possible to add a "Daily Score - All Time" graph to the Dashboard page? We currently have "Pairs Returned - All Time" but it's a little hard to make a comparison on that basis since the graph is dominated by the occasions when we had large amounts of resources crunching very small tests. A score graph could tell us quite handily whether this was truly the biggest rally in terms of what people brought to it.

I'm sure this one would be the tops in terms of total score but it's an apples to oranges comparison. Those rallies were only 2 or 3 days. I'm fairly confident that our very first rally in early 2009 had the most score processed per day.

[Mini-Geek]

Originally Posted by gd_barnes

I'm fairly confident that our very first rally in early 2009 had the most score processed per day.

This would definitely be the thing to count it by. So can we get some exact numbers for this for each rally to see where this one ranked?

[mdettweiler]

Well, as it turned out, adding the Total Pair Score chart wasn't too hard--I was able to do it myself pretty easily. So, now we have a "Total Pair Score - All Time" graph on the dashboard page right under the preexisting "Pairs Returned - All Time" graph. This new graph shows, essentially, the computational effort put in on a given day. (Note that it can be thrown off a little by manual results, which we import into the DB as entire ranges listed under the day that they were reported complete. However, aside from a few large spikes back in 2008 which I suspect are the large expanses of "Unknown" results we still have yet to clean up, I don't think this is throwing anything off too much. From 2009 onward, all the biggest spikes are the rallies. In fact, this issue would affect the Pairs Returned graph as well, just not quite as distinctly. You may notice that that graph has its biggest spike in December 2009--that was a large dump of manual 12th Drive results from Bruce.)

Looking at the new graph, it becomes immediately clear that this latest rally did indeed represent the largest amount of computational power crunching for NPLB at any given time. Even the early 2009 rallies and the Free-DC "Project of the Month" bonanza that came concurrently don't come close to this last rally, or even the one before it, in terms of daily output.

(Again, ignore any large spikes from 2008...I'm pretty sure those are "Unknown" results and thus they should disappear when we get those cleaned up. Any real rallies in 2008 would have had much smaller spikes.)

[kar_bon]

Originally Posted by mdettweiler

(Again, ignore any large spikes from 2008...I'm pretty sure those are "Unknown" results and thus they should disappear when we get those cleaned up. Any real rallies in 2008 would have had much smaller spikes.)

Hint:
In July 2009 I've made a graph of pairs returned to our LLRnet-servers from start of NPLB. The races are marked.