distributed.net OGR-25 ending soon?
I've been collecting the "percentage completed" data of distributed.net's OGR-25 project since August 6th, 2007. At that date, the ratio was 65096%. I even got one older date, 53.01% on May 13th, 2007. I'm comparing this value with the current one (76.32 % on December 22nd) and doing some calculations:
First, the expected date of completion currently is August 2nd, 2008. This value has risen almost ever since I was tracking this data. Over the time, our average daily completion ratio was at 0.105 % of the whole keyspace (of OGR-25 phase 2). For this december, the ratio rather is below 0.04 %. It seems the project is getting slower and slower. This is not true though for computation power. The project gets more and more power over the time. Stats currently state we did 322 Gnodes/sec yesterday while doing only 178 Gnodes/sec on average. The 322 are an arbitraty value but the trend is clear: more computation power but less progress.
How comes? Well, we compare apples with oranges here: the Gnodes/sec value really expresses computation power. The percentage completed value does not! It expresses how many stubs we have completed. One stub contains an arbitraty number of nodes and it is unknown beforehand how big it is. What currently happens is that the stubs get bigger and bigger (on average). So, while our computers get faster, the project's progress gets slower.
Some people promised an increase in progress near the end. In spring it was even rumored that the project could end this year. Well, doesn't look like this right now. Even though we got faster, the phase 2 (of 2) is only three quarters completed. We could still experience some faster progress in the future, but there also might be another problem: Some ranges of huge stubs still sit on the server and will be delivered anytime soon.
For the stats, all this has no influence. The stats express Gnodes and thus the real computational power of your machines. Only the percentage completed value of the whole project is biased by the quite uknown node-to-stub ratio.
I'll report again next year about our progress.
Daily updated stats: distributed.net OGR-25 stats
First, the expected date of completion currently is August 2nd, 2008. This value has risen almost ever since I was tracking this data. Over the time, our average daily completion ratio was at 0.105 % of the whole keyspace (of OGR-25 phase 2). For this december, the ratio rather is below 0.04 %. It seems the project is getting slower and slower. This is not true though for computation power. The project gets more and more power over the time. Stats currently state we did 322 Gnodes/sec yesterday while doing only 178 Gnodes/sec on average. The 322 are an arbitraty value but the trend is clear: more computation power but less progress.
How comes? Well, we compare apples with oranges here: the Gnodes/sec value really expresses computation power. The percentage completed value does not! It expresses how many stubs we have completed. One stub contains an arbitraty number of nodes and it is unknown beforehand how big it is. What currently happens is that the stubs get bigger and bigger (on average). So, while our computers get faster, the project's progress gets slower.
Some people promised an increase in progress near the end. In spring it was even rumored that the project could end this year. Well, doesn't look like this right now. Even though we got faster, the phase 2 (of 2) is only three quarters completed. We could still experience some faster progress in the future, but there also might be another problem: Some ranges of huge stubs still sit on the server and will be delivered anytime soon.
For the stats, all this has no influence. The stats express Gnodes and thus the real computational power of your machines. Only the percentage completed value of the whole project is biased by the quite uknown node-to-stub ratio.
I'll report again next year about our progress.
Daily updated stats: distributed.net OGR-25 stats
Kommentare