[ENet-discuss] Congestion Avoidance and Control
Aaron Boxer
aboxer at oanda.com
Sun Mar 8 19:24:57 PDT 2009
Thanks, Lee. Very interesting article. The results indicate that ENet
has bandwidth requirements very similiar to various
TCP variations, but it's packet latency is lower. So, it still might be
of use to me, but not to reduce bandwidth.
Thanks again,
Aaron
Lee Salzman wrote:
> Someone had done a paper on network latency in games, and they used
> ENet as one of the
> test libraries. It had some nice bar graphs in it, although I don't
> remember the exact contents
> of what it said anymore (it's been a while since I read it):
>
> http://heim.ifi.uio.no/~paalh/publications/files/netgames2007.pdf
>
> Lee
>
>
> Erik Beran wrote:
>> I'm quite interested in these results as well. I saw there was
>> someone using for an xbox/360 game a while ago and I know their TCR
>> (technical certification requirements) test against bad network
>> conditions. Bad network conditions being relatively low bandwidth,
>> high latency and a certain percentage of packet loss over time.
>>
>> Could that person mention at all if there was any problems going
>> through cert/tcr testing in respects to network conditions?
>>
>> /Erik B.
>>
>> On Fri, Mar 6, 2009 at 4:42 AM, Aaron Boxer <aboxer at oanda.com
>> <mailto:aboxer at oanda.com>> wrote:
>>
>> Great. This sounds like the perfect library for my application.
>>
>> I am streaming currency rates using TCP at the moment; the client
>> is at an arbitrary
>> IP address on the internet. To maintain "real time" streaming,
>> the packet size is small,
>> so the TCP header size / packet size is quite large. The
>> connection must be reliable.
>>
>> It looks like ENet would give me a better header/payload ratio.
>>
>> Thanks!
>>
>> Aaron
>>
>>
>> Lee Salzman wrote:
>>
>> The library was developed and tuned for 56K modem usage,
>> perhaps which is
>> sort of a joke in this day and age. :)
>>
>> There is no particular roadmap for new features. The library
>> has just grown stuff
>> as Sauerbraten has needed the features. Thus, it tends to be
>> rather stable over
>> time, and it's even pretty safe to fork the code or do
>> whatever with it rather than
>> treat it as a normal library because of that.
>>
>> Lee
>>
>> Aaron Boxer wrote:
>>
>> Thanks. How much testing has been done under high packet
>> loss?
>>
>> Also, in general, what are the known issues with the
>> library?
>>
>> Is there a roadmap for new features?
>>
>>
>> Thanks again
>>
>>
>>
>> Lee Salzman wrote:
>>
>> It uses pings or reliable packets to measure the mean
>> round trip time for the connection.
>> Any excessive deviation from this causes a throttle to
>> oscillate up or down, which regulates
>> how often ENet will drop unreliable packets to reduce
>> bandwidth usage.
>>
>> Lee
>>
>>
>> _______________________________________________
>> ENet-discuss mailing list
>> ENet-discuss at cubik.org <mailto:ENet-discuss at cubik.org>
>> http://lists.cubik.org/mailman/listinfo/enet-discuss
>>
>> _______________________________________________
>> ENet-discuss mailing list
>> ENet-discuss at cubik.org <mailto:ENet-discuss at cubik.org>
>> http://lists.cubik.org/mailman/listinfo/enet-discuss
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> ENet-discuss mailing list
>> ENet-discuss at cubik.org
>> http://lists.cubik.org/mailman/listinfo/enet-discuss
>>
>
> _______________________________________________
> ENet-discuss mailing list
> ENet-discuss at cubik.org
> http://lists.cubik.org/mailman/listinfo/enet-discuss
More information about the ENet-discuss
mailing list