It looks like 1242+1243 have made a significant difference. I wanted to 
quantify it before announcing them on the website.

I did 3 manual bootstrapping tests - create a new node from the installer, see 
how long it takes to reach different numbers of peers. These show 3 peers in 
under 30 seconds, 5 in around 1 minute and 10 in around 3 minutes. These are 
only accurate to 10 seconds or so due to manual timing, and in all cases I 
started it fetching various sites etc, so there is both path folding and 
announcement going on - as there would be on a newbie node usually. Note that 
the peers limit was 14, so it stops announcing at 7 peers, hence it taking a 
relatively long time to get to 10 peers.

Then I ran the automated bootstrap tester 5 times in succession. This announces 
and doesn't path fold, or doesn't path fold much. I got the following results 
(time to 6 peers; it announces right up to that point): 0:41, 8:22, 2:52, 0:27, 

It also ran this morning and got a result of 3:12. This is not exceptional. By 

August: 2:06
September: 2:15
October: 2:46
November: 2:51
December: 2:59
January: 2:47
February: 3:51
March: 3:53 (not including today)

As you can see the situation has got worse over time. :|

Anyway, the one that ran this morning was running at around 4AM GMT, it is 
possible that many nodes had not upgraded at that point etc, IMHO it is not 
very significant.

So what of the outliers in the group of 5 tests? It looks like they might be 
related to local networking issues. First, running them in quick succession 
means we will be bombarded by handshakes from previous instances. Second, this 
is running on my system which is a seednode with relatively low bandwidth and 
apache as well.

So I turned off apache and further throttled my seednode, and introduced a 6 
minute delay between each sample, and I got:

This is all rather confusing! Clearly my system is not the best place to be 
measuring this from.

Attachment: signature.asc
Description: This is a digitally signed message part.

Support mailing list
Unsubscribe at

Reply via email to