Example: Count all IP addresses close to a target key #33
Conversation
|
Added a histogram at the end of the lookups. This gives us insights into the probability a nodes is found within a given search. The graph is not the easiest to read. I am constrained by the library. Nonetheless, it shows the distribution quite well. Here is short explanation on how you can read this histogram: Default Bootstrap Nodes - 23 IPs found - 55 lookups - Example 1It's almost always the 17+1=18 same nodes. One node has a 50% chance and then there is a long tail. Default Bootstrap Nodes - 21 IPs found - 51 lookups - Example 2Similar to the first example. 17+1+1=19 nodes are almost always found. Additional 2 nodes are found Randomized Bootstrap Nodes - 50 IPs found - 50 lookups - Example 3Randomizing bootstrap nodes increases the sloppyness heavily. 10+2+3=15 nodes are hit with a chance of >75% for each lookup. 1+2+3+1+2=9 nodes have an average chance of ~40% of being hit and then there is a long tail of 23+3=26 nodes that might be hit occasionally. Randomized Bootstrap Nodes - 44 IPs found - 54 lookups - Example 4The same 11+6=17 nodes are basically hit all the time. 3+1+1=5 nodes are hit every ~50% and the long tail of 4+18=22 nodes are hit rarely. I hope this gives enough insights into how sloppy the algorithm is. Randomizing bootstrap nodes definitely adds randomness and should be added as an option in the library. |
|
All checks pass. Ready to merge IMO. |
Adds an example to count all unique IP address around a target key. Used to test how sloppy the algorithm is in finding a key.
The algorithm
Random Boostrap Nodes 1
Random Boostrap Nodes 2
Default Boostrap Nodes 1
Default Boostrap Nodes 2
Findings: