nvm, I think I solved this, the issue was in MySQL overhead I think.

Tuned a lot MySQL InnoDB storage + used INSERT DELAYED for sent_sms table
(this table is in MyISAM for now)

And btw, now bearerbox spreads load better:


smsc[smsc]SMPP:smpphost.com:2775/2775:smpplogin:NULL (online 2554s, rcvd:
sms 0 (0.00,0.00,0.00) / dlr 2497 (0.00,0.01,0.98), sent: sms 553
(0.00,0.00,0.22) / dlr 0 (0.00,0.00,0.00), failed 0
, queued 0 msgs)
smsc[smsc]SMPP:smpphost.com:2775/2775:smpplogin:NULL (online 2554s, rcvd:
sms 0 (0.00,0.00,0.00) / dlr 2205 (0.03,0.01,0.86), sent: sms 252
(0.00,0.00,0.10) / dlr 0 (0.00,0.00,0.00), failed 0
, queued 0 msgs)
smsc[smsc]SMPP:smpphost.com:2775/2775:smpplogin:NULL (online 2554s, rcvd:
sms 0 (0.00,0.00,0.00) / dlr 2042 (0.01,0.01,0.80), sent: sms 210
(0.00,0.00,0.08) / dlr 0 (0.00,0.00,0.00), failed 0
, queued 0 msgs)
smsc[smsc]SMPP:smpphost.com:2775/2775:smpplogin:NULL (online 2554s, rcvd:
sms 0 (0.00,0.00,0.00) / dlr 2201 (0.00,0.01,0.86), sent: sms 213
(0.00,0.00,0.08) / dlr 0 (0.00,0.00,0.00), failed 0
, queued 0 msgs)

I've tested also speed of sending sms via smsbox without DLR its so
awesome. I guess I need a new server with better hardware to get speeds
like this on MySQL storage.

2012/4/4 Kyriacos/Netsmart <[email protected]>

>  Hi,
>
> I believe the "load balance" happens only in the sence that if one bind is
> busy sending an SMS and a new request comes, instead of queuing it, it goes
> to the second bind of the same name, rather than doing "round-robin" load
> balancing which seems to be what most people expect/assume. If your send
> rate is fairly low, then most messages will leave via the 1st bind.
>
> Regards,
> Kyriacos Sakkas
>
>
> On 04/04/2012 13:58, spameden wrote:
>
> I'm using latest kannel from SVN with sqlbox in scheme:
> Box connections: smsbox:sqlbox, IP 127.0.0.1 (0 queued), (on-line 0d 0h
> 13m 12s)
>
> I'm using multiple connections to the same SMSC with the same login (my
> smsc allows up to 4 concurrent connections under the same login).
>
> I have configured 2 connections for sending SMS and 2 connections for
> receiving DLRs.
>
> I noticed from kannel's status kannel isn't load balancing properly
> between them:
>
>       mysmsc[smsc]    SMPP:smpp-provider.com:2775/0:login:NULL (online
> 623s, rcvd: sms 0 (0.00,0.00,0.00) / dlr 2559 (4.29,4.21,4.11), sent: sms
> 2559 (4.29,4.21,4.11) / dlr 0
>    (0.00,0.00,0.00), failed 0, queued 0 msgs)
>        mysmsc[smsc]    SMPP:smpp-provider.com:2775/0:login:NULL (online
> 623s, rcvd: sms 0 (0.00,0.00,0.00) / dlr 234 (0.40,0.38,0.38), sent: sms
> 234 (0.40,0.38,0.38) / dlr 0
>    (0.00,0.00,0.00), failed 0, queued 0 msgs)
>
> dlr's are here because I'm using dlr_mask=31 to get also state=8.
>
> You can clearly see that number of sent sms are not the same and differs
> in about 10 times or more.
>
> I can't use different smsc_id's because in this case messages won't be
> matched against dlr table.
>
> I can post relevant bits from configuration as well if needed..
>
>

Reply via email to