> TCPDUMP ANALYSIS OF INTERBASE QUERY RUN:
> ----------------------------------------
>
> query = "select count(*), 'RESULT' from test"
>
> Start
> client -> IB, 12 bytes
> + 065 ms
> IB -> client, 32 bytes
> + 000 ms
> client -> IB, 8 bytes
> + 050 ms
> IB -> client, 32 bytes
> + 000 ms
> client -> IB, 80 bytes with the query text
> + 059 ms
> IB -> client, 64 bytes
> + 000 ms
> client -> IB, 24 bytes
> + 049 ms
> IB -> client, 40 bytes
> + 000 ms
> client -> IB, 32 bytes
> + 083 ms
> IB -> client, 236 bytes, most of it is zero and there is a text
> "COUNT" twice
> + 000 ms
> client -> IB, 32 bytes
> + 110 ms
> IB -> client, 408 bytes, most of it is zero and there is a text
> "COUNT" twice
> + 000 ms
> client -> IB, 24 bytes
> + 048 ms
> IB -> client, 32 bytes
> + 000 ms
> client -> IB, 28 bytes
> + 058 ms
> IB -> client, 32 bytes
> + 000 ms
> client -> IB, 40 bytes
> + 084 ms
> IB -> client, 40 bytes with the text "RESULT"
>
> The total query time is 606 ms.
>
>
> ----------------------------------------------------------------
>

And additionally:

When I do remote backup (using service manager) with verbose output 30 MEGABYTES data was transmitted to the client who made the backup. The database was 1.1M, its backup is 160 k.
I used IBConsole.
What the hell this is? Do we really need the huge amount zeros to be transmitted?

I checked the Microsoft SQL server protocol. It has the same result as the MySQL and PostgreSQL. Only the Interbase is slow.

Viktor