John F Hall wrote:

In article <3n3bh1pclpaeonike8ln9ictfjb4m5uvua@4ax.com>, f/fgeorge  <*> wrote:

<fredclark@consltec.demon.co.uk> wrote:

[...]

(and are crunching). In other cases no downloads have been achieved, so
no crunching.
This is despite being booted up and running online (ADSL) for several
hours.

Because there are over 100,000 of us with over 400,000 computers ALL
trying to connect AT THE SAME TIME!

[...]

That's because Berkeley seem to have programmed a "Denial of Service"
attack against themselves.  It's clear their "backing off" stategy is
seriously flawed.  I hope they will rethink it.

I wonder if their exponential backoff works only for the initial contact with the servers. This would mean that it works well for when the servers are off, but it will not work well if congestion or connections limits problems drop the connection during later handshaking.

Do they use one TCP connection for the entire communication? Or must new connections be created to complete the transfer process?

Or might they just be running out of resource on their backend stuff?


[...]

So, what's wrong?  The main problem seems to be each request backs off
*independently*.  I would have thought that the uploads and downloads
should each be backed off as a *group*. ...

Good idea.

The second problem I see is that uploads and downloads seem to be aiming
at the same FQDN (setiboincdata...).  Surely there should be two FQDNs

[ or more for load balancing ]

I think Boinc is already intended to support that.


Regards,
Martin