On Fri, Oct 21, 2011 at 2:31 PM, Alex Karasulu <akaras...@apache.org> wrote:
> On Fri, Oct 21, 2011 at 2:12 PM, Emmanuel Lecharny <elecha...@gmail.com>wrote:
>
>> On 10/21/11 12:56 PM, Zigor Salvador wrote:
>>
>>> Comments after using Wireshark to sniff on the packets being sent:
>>>
>>> On 20 Oct 2011, at 16:10, Emmanuel Lecharny wrote:
>>>
>>> Each message sent and received on ethernet will use 1500 bytes, even if
>>>> you transmit only one byte of data.
>>>>
>>> This is not correct.
>>>
>>> Using MINA's "no delay" option and sending around 439 bytes of data the
>>> Ethernet frame captured by Wireshark reports a total frame length of 505
>>> bytes.
>>>
>>
>> That does not change the fact that Ethernet frames are 1500 bytes long...
>>
>>
>>> (505 = 6 (MAC) + 6 (MAC) + 2 (Type) + 20 (IP header) + 32 (TCP header) +
>>> 439 (data))
>>>
>>> If I enable Nagle's algorithm, several messages get crammed into a single
>>> ethernet frame (up to the aforementioned MTU value).
>>>
>> Nagle algorithm is not a good idea for small messages, as you will have a
>> delay between each messages, the system waiting for some more bytes to fill
>> the PDU as much as it can before sending it to the client.
>>
>> It's usefull when sending big data, though.
>>
>>
> Wondering if MINA can get a little smarter. Just a brain dump here .... what
> if you know according to the protocol that you're going to get a certain
> train of responses. Maybe there's a way to buffer and dump to fill to MTU
> capacity?
>
> --
> Best Regards,
> -- Alex
>
It's not a kernel level matter ? to fill the packet to MTU limit using Naggle ?
