Contents
Introduction
One of the longest-running web sites dedicated to distributed computing is distributed.net. Originally known for brute-forcing the key for various encryption schemes, such as RC5-56, their current primary project is almost certainly OGR-27. The OGR-27 project seems to find the optimal Golomb ruler of the twenty-seventh order. While previous OGR projects merely confirmed the optimality of previously known rulers, OGR-27 apparently has a good chance of actually discovering a new optimum ruler.
In order to construct this page and the statistics on it, I collect data from a couple of sources:
- The primary source of data regarding the current status of the OGR-27 project is the official OGR-27 status page, which provides information on the number of remaining stubs in each stubspace.
- In order to collect even more detailed data for the stubs in each stubspace, I am also analyzing the logs from my personal proxy (dc.calindora.com:2064), which is open for public use.
The results of my various analyses are presented below.
Projected Completion
While I would like to believe that this section uses relatively sound mathematical and statistical principles to make its projections, it is nonetheless based on incomplete data, and as a result, the projections may be wildly inaccurate. However, they can provide a rough guess about when OGR-27 might come to a close. Just don't send me massive amounts of hate mail if they turn out to be twenty years off.
All projections are updated dynamically based on the stubspace statistics you will find later in this document. As such, even though I do list the previous seven days worth of projections, they are constantly being revised to reflect the new knowledge about the distribution of stubspace sizes. The primary purpose of displaying the most seven recent projections is to allow you to identify any recent trends in computation speed.
Unfortunately, I am making some questionable assumptions in calculating these projections:
- I assume that future stub sizes (within a given stubspace) will be distributed in a similar fashion as stubs that have previously been encountered. I am not entirely sure this assumption is valid, as it depends on whether or not we are traversing the stubspace in a linear or random order.
- At this time, I am also assuming that the computation rate will double every two years, in an attempt to account for increases in computer speed as time goes on. However, since participants may join or leave the project at any time, this can be nothing more than a rough approximation. In time, I may adjust this rate to reflect actual project data.
For the curious among you, the average number of Gnodes per day is calculated as an exponentially-weighted average of the with a half-life of seven days, based entirely on my estimated number of Gnodes remaining.
Most Recent Seven Projections
| Date | Remaining | Avg. Gnodes / day | Projected End Date | |
|---|---|---|---|---|
| Stubs | Gnodes (estimated) | |||
| 2010-07-24 | 515,747,941 | 56,844,547,355.078 | 21,979,165.227 | 2014-02-20 |
| 2010-07-25 | 515,517,705 | 56,819,171,274.701 | 22,299,413.940 | 2014-02-10 |
| 2010-07-26 | 515,517,705 | 56,819,171,274.701 | 20,197,106.905 | 2014-04-27 |
| 2010-07-27 | 514,986,599 | 56,760,633,997.540 | 23,811,677.668 | 2013-12-24 |
| 2010-07-28 | 514,727,159 | 56,732,039,119.706 | 24,262,620.253 | 2013-12-11 |
| 2010-07-29 | 514,423,018 | 56,698,517,400.854 | 25,135,534.141 | 2013-11-16 |
| 2010-07-30 | 514,086,937 | 56,661,475,328.940 | 26,258,038.893 | 2013-10-17 |
Historical Projections
Since the projections are dynamically recalculated based on changing data about the stubspaces, the older projections would no longer reflect what was actually projected at the time. However, historical projections remain useful as a tool to see how accurate the projections (especially the earlier ones) turn out to be. To aid in this goal, I am manually recording the latest projection every three months and posting it in the following table:
| Date | Remaining Stubs | Projected End Date |
|---|---|---|
| 2009-06-30 | 591,935,532 | 2015-10-26 |
| 2009-09-30 | 578,373,169 | 2014-02-19 |
| 2009-12-31 | 562,327,523 | 2014-01-10 |
| 2010-03-31 | 541,258,984 | 2013-11-01 |
| 2010-06-30 | 520,385,618 | 2013-11-17 |
At this point, the end date seems to have stabilized somewhat. Of course, it's still just an estimate.
Stubspace Details
OGR-27 has a relatively simple list of stubspace definitions. There are four different stubspaces, defined solely by the number of diffs in each stub. More information is available on the aforementioned OGR-27 status page. The following data is based exclusively on the stubs that have passed through my personal proxy. I would like to think that this forms a representative sample, but there is, of course, no guarantee.
First, here are the vital statistics for each stubspace:
| Stubspace | Seen Stubs | Sizes | ||
|---|---|---|---|---|
| Mean | Minimum | Maximum | ||
| 1 | 6 | 336.967 | 322.977 | 351.851 |
| 2 | 204 | 244.238 | 45.638 | 602.371 |
| 3 | 11775 | 124.019 | 2.215 | 747.039 |
| 4 | 24392 | 110.218 | 0.972 | 1,243.850 |
For the more visually inclined, the following graph displays the distribution of stub sizes for each stubspace:
