The small ice cream shop Tiny Treats might have only one network-connected smart cash register. Any request from that public IP likely comes, therefore, from that single client (unless they have not secured their WiFi network...). There is a 1:1 ratio of clients to source IP addresses from FortiWeb’s perspective.

Down the street, Giant Gelato, which distributes ice cream to eight provinces, might have a LAN for the entire staff of 250 people, each with one or more computers. Requests that come from the Giants Gelato office’s public IP therefore may actually originate from many possible clients, and therefore normally could be much more frequent. However, like many offices, the LAN uses source IP network address translation (SNAT) at the point that it links to the Internet. As a result, from FortiWeb’s perspective, the private network address of each client is impossible to know: it only knows the single public IP address of Giant Gelato’s router. So there is a single source IP address for Giant Gelato. However, there is a 250:1 ratio of clients to the source IP address.

This is a big proportionate difference. While a low rate limit might seem generous to Tiny Treats, Giant Gelato would be unhappy if you applied the same rate limit to its IP address.

Let’s say that both companies need access to the same ice cream inventory web application: Tiny Treats buys from Giant Gelato. Each view in the application contains the page itself, but also up to 15 images of ice cream, 3 external JavaScripts, and an external CSS style sheet, for a total of 20 HTTP requests in order to produce each view.

40 requests per second then might be more than adequate for Tiny Treats: the clerk could page through the inventory twice every second, if she wanted to.

But for Giant Gelato, its clients would frequently see completely or half-broken views: some images or CSS would be missing, or page requests denied the first or second time, because some other clients on Giant Gelato’s LAN had already consumed the 40 requests allowed to it per second of time. Normal use would be impossible.

To be practical, then, you would not base your rate limiting solely on the source IP address of requests. Instead, you would want dual thresholds:

You could enable Shared IP so that FortiWeb could know to permit more requests per second from Giant Gelato than from Tiny Treats. Because Giant Gelato’s ID fields would not usually be continuous as a single client’s usually would be, FortiWeb could then apply a different, higher limit.