On Sat, 29 Oct 2005, Bob wrote:
> Roman Bednarek <roman at ...> writes:
>
>> On Fri, 28 Oct 2005, Matthew Toseland wrote:
>>
>>> It *should* work. It is believed to work. Do you have an
>>> inputBandwidthLimit set? That might be helpful - incoming traffic
>>> requires acknowledgement via outgoing packets.
>>>
>> No, I do not have inputBandwidthLimit set (my ADSL has much bigger
>> input than output), but now when I set it nothing has changed. I run it on
>> linux, if it doeas matter. On servlet/nodeinfo/performance/general page I
>> can read:
>> Current upstream bandwidth usage 208 bytes/second (5,1%)
>> and at the same time about 18KB on iptraf monitor(with only freenet
>> running).
>> I had to stop freenet node, to write that answer in pine, full
>> bandwidth was used and I even could not type. (not always is so bad, most
>> of the time something free is left).
>>
>> Roman
>
> Hmm, well that's odd. Output limiting is not accurate, and there can be a lag
> of
> up to 10 minutes before fred notices changes to the conf file, but it seems to
> basically work in my experience. Some disparity between fred's usage report
> and
> iptraf's could be explained by instantaneous vs. long period sampling, but
> can't
> explain a limit of 4k apparently maxing out your upstream.
>
> Could you post your outputBandwdithLimit line exactly as it appears in
> freenet.conf / freenet.ini, and maybe the immediately surrounding entries? I
> suspect it's not doing anything at all because it's somehow malformed, thus
> letting freenet run unlimited.
>
> Bob
>
The config option is specified correctly, I see that it is changing
bandwidth usage. Through trial and error and discovered that setting limit
to 2KB is acceptable for my upload bandwidth, it takes about 10-15KB. In
report "Current upstream bandwidth usage" is now a little bigger, around
specified 2KB, it is not the same what iptraf shows, but is better than
previous 0.2 KB.
So, the limit is not exact, but is working, and I was able to tune it
to my needs.
Long time ago, before Asynchronyous IO limiting was accurate, probably
limiting with nio is not always working as expected.
Roman
