"FeedBurner Saved My Bandwidth!"

The Wired article "Will RSS Readers Clog the Web?" highlights the growing concern over the aggregate bandwidth consumption of polling news readers. Trying to address this problem was actually one of the original motivations behind FeedBurner! Read on to see some of the original design documents that show how FeedBurner helps publishers manage their bandwidth.

For this example, let's say there are 50,000 RSS clients out there that are watching an RSS feed for updates on someone's blog. We can do a more detailed analysis later that looks closer at the nature of these requests (conditional HTTP GET vs. HTTP GET and typical payload sizes), but let's keep things simple now. Here's what the situation looks like today for these clients checking every 30 minutes for 16 hours a day:

This blog will have 1.6MM requests each day, just from clients checking for updates. Now, let's bring FeedBurner into the picture. The individual at somewhere.com now gives out the FeedBurner post-processed feed URL instead of his own URL. We have the infrastructure to handle the requests and we do some intelligent caching, so now it looks like this:

We provide the blog guy all the stats he wants and he doesn't worry about his ISP yelling at him for consuming too much bandwidth because now instead of 1.6MM requests for the RSS feed, he's only getting 288. (As an aside, if the publisher pings FeedBurner, we can avoid polling altogether.)

This has improved the situation for the RSS publisher, but still results in a lot of polling. Now let's say we work with the RSS clients to implement a pub/sub model based upon something like the XML-RPC Ping protocol or even something like BitTorrent for feeds. Now we start turning those pulls into pushes:

Now, obviously, the devil is in the details in figuring out how that would really work with clients (behind firewalls, or disconnected, etc.), but there's an incentive for a client to actually implement it so they get the update message essentially instantly instead of waiting for the next polling period.

This is how FeedBurner works today and a glimpse into what we're planning for the future.