[tor-talk] Avg packaged cell fullness - notice in logs

[Nick Mathewson]

On Mon, Dec 9, 2013 at 9:44 AM, Jon <torance.ca at gmail.com> wrote:
> I have recently noticed since  I have gone from an exit relay to a non-
> exit relay.
>
> In the logs, I have noticed that the Avg packaged cell fullness is showing
> 26% - 50% ( rounded off ) on the non-exit relay.

A non-exit relay should be packaging very few cells ; if it's not a
directory cache, what it packages will be more or less entirely
directory requests, which are pretty short. I wouldn't be too worried,
unless you happen to be a directory cache, in which case we should
look into some of the ideas we had for improving cell packaging.

> When i was running the exit relay the Avg packaged cell fullness was
> showing 93%-98%.
>
> Also the TLS write overhead on the non-exit relay is showing 8% - 13%. The
> exit relay was showing 6% - 8%.

This one, I'm not so sure about. Is this value typical for folks?

> Is this normal?  The only thing that has changed is going from exit relay
> to a non exit relay.
>
Best wishes,
-- 
Nick