Re: [PATCH] PostgreSQL DLR speedup

Alexander Malysh schrieb:
> Hi,
>
> Andreas , this is not true what you say. Mysql use LIMIT 1 in delete 
> and update. Postgres should use oid because there is not limit support.
>
> It's there because e.g. on EMI we could delete/update to much rows.
Yes, because the implementation is broken for EMI anyways.

The combination of SMSC and Timestamp *IS NOT* unique, the combination of
SMSC, Timestamp and MSISDN *IS* - at least for my operator.

As the API obviously passes the MSISDN, but the SQL statement makes no use
of it, there seemed to be a reason not to include the MSISDN in the 
query - but which?

I just did a SELECT on my DB of roughly 60.000 open DLRs:

mysql> select ts,count(*) as a from mydlr group by ts order by a desc
limit 10;
+--------------+----+
| ts           | a  |
+--------------+----+
| 091106080850 | 60 |
| 091106082414 | 47 |
| 091106081849 | 47 |
| 091106082234 | 46 |
| 091106082059 | 45 |
| 091106082227 | 44 |
| 091106081909 | 44 |
| 091106082601 | 44 |
| 091106080948 | 42 |
| 091106082334 | 42 |
+--------------+----+


Cheers,


Georg