Re: [mosquitto-dev] [RSMB] Strange memory behavior

[Mohamed HAMZAOUI <requinham@xxxxxxxxx>]

Hi Ian,

Thank you for your response.
I'm not using RSMB for MQTT-SN but for MQTT. I have tested Eclipse
Mosquitto before but I have detected a strage memory behavior :
The test :

UBUNTU Trusty <============IP/ETH============> Raspberry PI3
Simple Publisher
         Eclipse Mosquitto 1.4 and simple Subscriber client

The result is to transferring message with 100MB  :

+-------------------+--------------------+-------------+
|                     | Memory Size | Time (s) |
+-------------------+--------------------+-------------+
| MQTT QoS-0 | 311752 KB    | 37          |
+-------------------+--------------------+-------------+
| MQTT QoS-1 | 311752         | 41          |
+-------------------+--------------------+-------------+
| MQTT QoS-2 | 311752         | 65          |
+-------------------+--------------------+-------------+

I don't understand exceeding 300 MB to transfer 100MB message size
between one publisher and one subscriber ! Then I tried to test RSMB
code but more problems.

Is the memory behavior is normal ? How do you qualify Eclipse
Mosquitto code of adapted to small and embedded systems ? Is there
another compiler option to fix this memory peak ?

Regards,
Mohamed

On Wed, Oct 12, 2016 at 6:07 PM, Ian Craggs
<icraggs@xxxxxxxxxxxxxxxxxxxxxxx> wrote:
> Hi Mohamed,
>
> yes RSMB does have that behaviour - the read buffer is increased to the size
> needed to read in that message, but never reduced.  You could raise an issue
> if it's something you'd like fixed.
>
> I assume that you are using RSMB for MQTT-SN?  We have a newly contributed
> MQTT-SN gateway which is currently undergoing testing.  Perhaps that would
> be a good replacement?
>
> Ian
>
> On 10/09/2016 11:02 PM, Mohamed HAMZAOUI wrote:
>
> Dear all,
>
> RSMB lead to be small broker but I have use RSMB to publish a "big" binary
> message (100MB) and the memory increases when the message was sent and no
> longer descends. Looks like the memory allocated for the message transfer
> was not as deallocated. The behavior is not cumulative: if published after
> the big message any other smaller message, memory size does not change but
> if we try to publish bigger one, the memory increases and do not fall. RSMB
> keeps memory allocated equal to the size of the largest forwarded message.
> Is this normal ? (Mosquitto don't have this strange behavior)
>
> Regards,
> /Mohamed
>
>
> _______________________________________________
> mosquitto-dev mailing list
> mosquitto-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or unsubscribe from
> this list, visit
> https://dev.eclipse.org/mailman/listinfo/mosquitto-dev
>
>
> --
> Ian Craggs
> icraggs@xxxxxxxxxx                 IBM United Kingdom
> Paho Project Lead; Committer on Mosquitto
>
>
> _______________________________________________
> mosquitto-dev mailing list
> mosquitto-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or unsubscribe from
> this list, visit
> https://dev.eclipse.org/mailman/listinfo/mosquitto-dev