Accurately count active users? (PHP + MySQL) - php

Question

Basically, I've been trying to get an accurate measure of how many active users my site has. I'm tracking sessions on the DB, and also doing some clean-up based on the USER_AGENT to remove bots/spiders/crawlers (very basic stuff though).
What I get is a very high number of sessions, even after removing duplicate sessions for the same IP. In fact the number of sessions is 10x larger than what Google Analytics and other User Tracking systems report, so there's gotta be something I'm doing wrong.
Here's part of the code where I keep track of sessions:
//Start session if not active already
if(!isset($_SESSION)){
session_start();
}
//Determine whether bot or browser
$bots = array(
'bot',
'crawler',
'yahoo',
'spider',
'google',
'$^'
);
$ua = $_SERVER['HTTP_USER_AGENT'];
$uam = preg_match('/'.implode('|', $bots).'/i', $ua);
$uatype = ($uam)?'bot':'browser';
//Add session to DB if new or expired
if ($_SESSION['renew'] < time() || !isset($_SESSION['renew'])) {
$_SESSION['renew'] = time() + 900; //15 minutes
$sql = "INSERT INTO " . SESSIONS . " (session_id, user_id,
renew, user_ip, type, useragent) VALUES ('" . session_id() . "',
'" . $myuser->get('user_id') . "', " . $_SESSION['renew'] . ",
'" . $myuser->get('ip') . "', '$uatype', '$ua')
ON DUPLICATE KEY
UPDATE renew = " . $_SESSION['renew'] . ",
user_id = '" . $myuser->get('user_id') . "'";
$site_db->query($sql);
$sql = 'DELETE FROM ' . SESSIONS . '
WHERE renew < ' . time();
$site_db->query($sql);
}
So, I'm storing the session info in DB and then renewing / expiring every 15 minutes as necessary.
If I then query the DB for sessions that are from browsers, with unique IP addresses, I get about 10x as many as JS systems are reporting.
Even if I try to reduce the number by only counting entries that have their first two octets of the IP address unique, I'm still way over.
Any tips on what I could be doing wrong on my session tracking?
The query to get the counts goes like this:
$sql = "SELECT count(DISTINCT(user_ip) FROM ".SESSIONS." WHERE type = 'browser'";
As a note, the bot trap is already discarding about 66% of the sessions. If I didn't do that, I would get close to 40x as many "active users" as GA reports.
Thanks.

Answer 1

Maybe storing those records and then do a query like this:
$sql = 'SELECT COUNT(*) FROM ' . SESSIONS . ' WHERE renew BETWEEN ' . $expire-900. ' AND ' .$expire;
And for performance reasons, let a cronjob do your deleting.
For example: Delete every record that expired an hour ago. And do that every half hour or something.