> So I can only expect 68 transaction per second or roughly
> 250,000 transactions per hour (I agree, row count is beside
> the point)? I will run a recalibrated test to see how far
> this gets me.

It appears a second opinion is in order.
I decided to perform a similar test to you: 4 million inserts, and timing
them.
Hardware:
P2 350MHz, 128MB ram, IDE disks, Firebird 1.0.2

Using the advice and feedback from your original email:
Don't commit on every insert, but commit every 10k or so
Do prepare the query in advance, otherwise Firebird will prepare it EVERY
SINGLE QUERY

SO I did the following:

--- Table Schema ---
create table tbl_test (
int_id NUMERIC(18,0) NOT NULL,
PRIMARY KEY(int_id)
);

So it's got an index to update(pkey), and a unique constraint to check(pkey)



--- PHP file --- (I don't have the money for commercial software) ;P
#!/usr/local/bin/php -e
<?php
$start = microtime();

// Connect to the DB
$conn = ibase_connect("twister:/u1/db/test.gdb","sysdba","masterkey");

// Prepare a statement
// (Cause firebird will do it for us for EVERY query if we don't)
$prep = ibase_prepare("INSERT INTO tbl_test VALUES (?)");

// Do the loop
echo "Inserting 4M records\n";
for($a = 0; $a < 4000000;$a++){
//Run the query
$res = ibase_execute($prep,$a);
if($a%10000 == 0 && $a > 0){
// WOrk out how fast were going
$smicro = substr($start,2,8);
$ssec = substr($start,11);
$begin = "$ssec.$smicro";

$finish = microtime();
$fmicro = substr($finish,2,8);
$fsec = substr($finish,11);
$end = "$fsec.$fmicro";
$perf = $end - $begin;
echo "$a records. Recs/Sec:".number_format($a/$perf,2)."\n";
ibase_commit_ret();
}
}

?>



--- Results ---
Inserting 4M records
10000 records. Recs/Sec:1,369.73
20000 records. Recs/Sec:1,387.93
30000 records. Recs/Sec:1,384.66
40000 records. Recs/Sec:1,378.88
50000 records. Recs/Sec:1,366.23
60000 records. Recs/Sec:1,356.47
70000 records. Recs/Sec:1,357.22
80000 records. Recs/Sec:1,358.60
90000 records. Recs/Sec:1,355.69
100000 records. Recs/Sec:1,353.22
110000 records. Recs/Sec:1,355.22
120000 records. Recs/Sec:1,357.30
...
360000 records. Recs/Sec:1,341.41
370000 records. Recs/Sec:1,341.66
380000 records. Recs/Sec:1,339.19
390000 records. Recs/Sec:1,339.83
400000 records. Recs/Sec:1,340.72
410000 records. Recs/Sec:1,341.28
420000 records. Recs/Sec:1,340.68
430000 records. Recs/Sec:1,340.47
440000 records. Recs/Sec:1,340.84
450000 records. Recs/Sec:1,341.69
460000 records. Recs/Sec:1,339.42
470000 records. Recs/Sec:1,339.61
480000 records. Recs/Sec:1,339.58
...
1110000 records. Recs/Sec:1,347.70
1120000 records. Recs/Sec:1,347.78
1130000 records. Recs/Sec:1,347.80
1140000 records. Recs/Sec:1,347.72
1150000 records. Recs/Sec:1,347.65
1160000 records. Recs/Sec:1,347.57
1170000 records. Recs/Sec:1,348.00
1180000 records. Recs/Sec:1,348.16
1190000 records. Recs/Sec:1,347.96
1200000 records. Recs/Sec:1,348.19
...you get the idea.


As you can see, it's quick, and it doesn't slow down.

Need a method to see if the table 'tbl_test' exists:
select first 1 1 from tbl_table;

Result: 1

Hope this helps

Nigel.