[ There is one other issue read below ]
Ok when I read the protocol document, insert seemed
like "not-propogating". And first replies that came to
my mails said that there was no key "advertising". So
I could not understand it earlier. Sorry. We learn new
thing every day :) Thanks everyone!
But the current scheme is still not %100 reliable. In
other networks (UUNET, Internet (FTP/HTTP), Novell,
USENET, etc.) the only reason for data unavailability
is "software or hardware" shortage. Network topology
never affects "data availability". Once you know the
address of data "data", you can always access it (if
there is no hw/sw shortage).
What I understand is:
- Freenet is efficient
there is no broadcast, ever
- Freenet is survivable
nodes are not considered
to be of equal power
- Freenet is NOT reliable
Whether you can access some "data"
is random (but VERY highly probable)
[ the other issue starts here ]
There is one other issue: the average storage space.
Currently it is very hard to access anything other
than snarfoo with even 5MBit bandwith (yes I currently
have 5Mbit badwitdh). So I did think a little bit
more...
The data availability ratio increases with the amount
of nodes that data is replicated. The default is
replication to HTL nodes, and the freenet simulation
got near %100 availability with the default HTL value.
In order for the avarage replication count to be
around HTL, the data storage must be:
S = key_count * avg_key_size * HTL / num_nodes
Since the current default storage space is ~95MB and I
assume that default insert HTL is 50, the avarage key
size must be
avg_key_size = num_nodes * 95MB / key_cnt / 50
(50 is the default for freesite insertion document)
I assume key_cnt / num_nodes (avg_key_cnt) to be 20.
Then we have:
avg_key_size = 95MB / 50 / 20
avg_key_size = 97K
(avg_key_size is not num_nodes / key_cnt so i did
divide it, don't get it wrong.)
So if every node is currently inserting an avarage
of 20 keys, we should have the avarage key size to be
97K
for very good reliability. But since everyone is
inserting their 4-8MB mp3s and 200-300MB iso's, we
have a "slow" and "less reliable" freenet.
(One may argue that larger keys have less life. But we
do not want keys larger than 100K disapper suddenly).
Since we cannot limit key size on users I assume we
need to change "default" storage size to several
hundreds for better freenet.
> Which is exactly what Freenet does. Inserts follow a
> similar path that requests do, so a node which is
> "close" to an inserted document will also be "close"
> for a request for that same docuement.
=====
Sukru Tikves,
Hacettepe University
Computer Engineering
Department, Ankara
http://piskare.cjb.net
__________________________________________________
Do You Yahoo!?
Listen to your Yahoo! Mail messages from any phone.
http://phone.yahoo.com
_______________________________________________
freenet-tech mailing list
[EMAIL PROTECTED]
http://lists.freenetproject.org/mailman/listinfo/tech
