Netty speed bottleneck

[Christian Migowski]

Hi,

On Wed, Jan 26, 2011 at 4:02 PM, frankdurden <frank.durden at gmail.com> wrote:
>
> Hi,
>
> I'm testing networking speed between two blades connected via 10Gbit
> ethernet. My test is very simple and uses Google's Protocol Buffers to send
> big messages between the machines. I'm getting a speed of ~35MBytes/sec
> which is not that impressive. This is a memory to memory test, no file IO.
> ttcp tells me that the speed should be at least 400MBytes/sec. So I'm
> wondering where the bottleneck is, is it how I use Netty, is it Netty itself
> or does it have to do with Java sockets. Can I do anything about this? Does
> the 'More New I/O API' address this in anyway?

Taken the information you've given us and given the fact that higher
speeds with Netty are possible and have been seen by many people, the
answers are:
1. It is how you're using Netty.
2. yes
3. maybe

regards,
christian!