[RC5] How to Interpret Keyspace Plots, please?

Dennis Lo dkslo at netvigator.com
Wed Jun 17 15:37:50 EDT 1998


Is it possible to modify the sequence / procedures of switching keyspace as
follow?

Assumption:
Keyspace 0x64, 0x66 and 0x6A are new keyspaces

1) If possible, D.Net distributes all (100%) keys in keyspace 0x64 before
D.Net starts to issue new keyspace.   (It will take around 30 days to issues
all keys in a keyspace)

2) If the master gets to < 3% of a subspace left and it looks like it will
run out while the master admin sleeps, the administrator should instruct the
server to issue THE NEXT (i.e. 0x66) keyspace.

3) ON THE FOLLOWING DAY, switch back to keyspace 0x64, issue all the
remaining un-issued keys until 0x64 reach 100%, then switch to keyspace
0x66. (Again, it will take around 30 days to issue "all" keys in 0x66)

4) If un-issue keyspace of 0x66 < 3% and it looks like it will run out while
the master admin sleeps, the administrator should switch the server to issue
THE NEXT (i.e. 0x68) keyspace.

5) ON THE FOLLOWING DAY, re-issue all the un-checked keys in keyspace 0x64
(At least 30 days have passed since the last issue of 0x64 keyspace)

6) When the re-issue of 0x64 is completed, switch to 0x66 and issue all the
remaining keys in 0x66.

7) When 6) is completed, switch to THE NEXT (i.e. 0x68) keyspace

8) Repeat procedures 1) to 7) for keyspace 0x6A, 0x6C......

I think the procedures mentioned above can make forecasting which keyspace
will be distributed much easier. Participants can go to
http://www.distributed.net/statistics/graphs/  and make a simple calculation
on when will D.Net re-issue the keyspace. Participants can adjust their
buffer size accordingly. Hence, it reduce the possibility of two
participants calculating same block of keys.

I understand that the proposed procedures will increase the work load of our
master administrator substantially, but to a certain extend it could solve
the problem of
"Lack of communication" raised by other participants in this email list.

I really hope that we could improve the speed of D.Net
----------------------------------------------------------------------------

>>We are issuing about 3% of a subspace / day , the master needs human
>intervention
>>to switch subspaces, so when the master gets to < 3% of a subspace left
>and it looks
>>like it will run out while the master admin sleeps, we switch to a
>different space. Rather
>>than going back the next day and finishing the issue we just leave the
>gap to be caught
>>up the next time that subspace is issued.
>>
>>Sent: 0x64:  98.03%
>>Checked: 0x64:  93.52%
>>Sent: 0x66:  81.45%
>>Checked: 0x66:  81.44%
>>Sent: 0x68:  94.54%
>>Checked: 0x68:  86.39%
>>Sent: 0x6a:  34.87%
>>Checked: 0x6a:  25.52%
>>
>>note that 5% of 0x64 is issued and not returned and 8% of 0x68
>>we should be continuing 0x6a in about 2 more days







--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list