Re: [GNUnet-developers] gnunet download problem

[Igor Wronsky]

On Wed, 26 Mar 2003, eric haumant wrote:

> To do so, I've inserted a 100Mbytes' file (it has taken 30 min on a P4
> 2.4GHz !!).

Inserted or indexed? In any case, that mostly depends on how the
physical disk is used by the dbmgr. CPU doesn't help here. When I
did some benchmarks on gdbm some time ago I noticed that atleast
inserts to gdbm database grew significantly slower and slower
as the database got bigger. :(

I haven't really benchmarked the mysql module in current gnunet
cvs, but I have a gut-feeling that it would scale better.

Anyway, content encryption or the merkle-tree construction
shouldn't be issue here, they're quite fast.

> When I asked for the file from another peer on the network, it starts
> downloading the file but after a few minutes (45min on average), it stops
> and there is no more activity between my gnunet's nodes. I have to stop and
> run gnunet-download once again for the download to continue.
> Is there any reason why this is happening and how can it be solved ?

Its a bug. Its not supposed to be happening. I've noticed it myself
when asking the file from the same node I inserted the file into, and
as all of the file can be fetched by restarting the dl enough times, its
hardly a database or content-gone-missing problem. ;) Unfortunately, it
means that its in the routing direction and I don't care to touch
that part (don't know it too well), and Christian is roaming
somewhere around the europe. Most likely the fix'll have to wait until
he gets back. A week tops, I'd think.

> For the moment, I've downloaded only 8Mbytes from the network in 1h30.

One thing certain is that atleast with the current implementation,
it will never be blazing fast, even if it worked correctly. A thing
worth mentioning anyway is that if you limit your upstream, that
will affect the download speed as well, as the node will drop
excess outgoing packets, even those containing your queries.


Igor