Question

For a web application I’m writing it’s going to be key to measure at a frequent interval (every minute or so) how many visitors are in a specific part of my application right now. Say my application has 10 subsections that are switched between using JavaScript and made persistent using #!/page8 style hash anchors. What would be the best way of doing this accurately?

My current plan is to just save the amount in a database, adding and subtracting every time a user clicks on the page or leaves the page. This doesn’t really seem like a very good solution to me though. Events to track page leaves could be cancelled due to closing browsers, people could be screwing with the data by running the API calls for visiting/leaving etc. It would be a lot of work to get reliable.

Edit: To specify, my application won’t have user accounts. Also I’m not just trying to log how many visitors my pages have had. I’m trying to accurately know how many visitors are on my page right now. Which means that I’ll need some reliable method of knowing then a visitor is no longer on my page. That’s where the problem lies. I just got a bright idea that I could have the JavaScript on their page notify my server every 10 seconds or so that they are still there, and have their visit be timed out and removed if they don’t. Would that work? Couldn’t this be a problem for my puny server if I were to get thousands of people using my app simultaneously?

Any thoughts?

---

_{I’m tagging this with a lot of common web development tags since it’s a very open question and any of these tags could contain the answer}

Answer 1

Depending on the server capacity you have you could employ recurring requests to a .php file that would keep a custom session track alive. But beware – this causes wast load on the server, so only go this way if you are aware of the consequences!

With session_start(); you could write a custom session identifier in a database table with fields like ID, ip_addr, session_track_id, timestamp. Don’t use the session_id itself as that might be a security issue. Update the timestamp in each .php you call.

Additionally create a custom .php where only the timestamp is updated. Call this via jQuery AJAX calls every 60 seconds (or whatever interval you need). You could employ some security to this step too to avoid direct API call you mentioned. For example issue a next request id that is returned in the AJAX calls and needs to be a parameter of the next one, if it’s not, the timestamp won’t be updated.

Create a cron that runs every 90 seconds (or whatever interval you need – but longer than in the AJAX calls obviously). This cron would look for entries in the table that are older than X seconds and delete them.

To find out the number of active users simply count the rows in the table.

There are additional security options that could be employed in this approach.

This would solve your situation – when a user would close his window, no AJAX calls would keed the session alive anymore and it would get deleted from the database after some seconds (90 in the scenario above).

However – consider that when you issue a request every 60 seconds and you have 1000 users online, it’s at least 60 000 requests per hour to you database! Strong optimalization is a must here…

This solution worked fine for me in one project however.