mina-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Zvika Gart <zvikag...@gmail.com>
Subject Re: Miserable performance results from MINA 1.1.7
Date Fri, 27 Nov 2009 15:19:23 GMT
>
> Keep in mind that the way you used MINA is really very specific. I

don't have such performance simply because I never transfer such big

messages. If you consider that messages that will be exchanged never

exceed a few kilobytes, then performances are also excellent.


The way I used MINA is the most basic. It should be performant. The uploaded
data can be FTP data, HTTP Form data, etc. The current MINA code is capping
the upload bandwidth.
In a scenario where the data only flows from client to server, even with
small messages you got a problem - the message rate is capped since the
window size is capped.

AFAIK, this is a configurable option, and if we don't configure it

before connecting the socket, then the RCV buffer size will never grow

bigger than 64K. So what should we do ? Set the buffer size to 2Gb ?.


> Also consider that I'm not 100% sure of the consequecnes of such a

change. Maybe it's totally harmless.


You don't need to configure anything. The OS itself grows/shrinks the
receive buffer & window size as needed. We saw that in the simple
implementations. I didn't configure the receive buffer and it (and the
receive window) grew to 2MB.


On Fri, Nov 27, 2009 at 4:34 PM, Emmanuel Lecharny <elecharny@apache.org>wrote:

> > First, I really appreciate your work on the framework, it's very easy to
> > work with and has really cool features, but I have to say that it is bad
> > that the out-of-the-box performance of MINA is so poor.
>
> Keep in mind that the way you used MINA is really very specific. I
> don't have such performance simply because I never transfer such big
> messages. If you consider that messages that will be exchanged never
> exceed a few kilobytes, then performances are also excellent.
>
> This simple
> > performance comparison test should always be verified throughout MINA
> > releases.
> > Second, why mass around with the receive buffer anyway if the user didn't
> > explicitly wants to? let the OS manage it...
>
> AFAIK, this is a configurable option, and if we don't configure it
> before connecting the socket, then the RCV buffer size will never grow
> bigger than 64K. So what should we do ? Set the buffer size to 2Gb ?.
>
> Also consider that I'm not 100% sure of the consequecnes of such a
> change. Maybe it's totally harmless.
>
>
> > Can you please release a patched version of MINA 1.1.7 with a fix?
>
> If you can provide a patch for 2.0§RC1, then we can fore sure apply
> it. I'm afraid that 1.1.7 is a dead branch at this point... Anyway,
> feel free to send the patch, I don't know if someone has some time to
> release a 1.1.8
>
>
> --
> Regards,
> Cordialement,
> Emmanuel Lécharny
> www.iktek.com
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message