Connection Limiting Overview
The Connection Limiting feature is intended to offer an additional layer of security and control when coupled with such SonicOS features as SYN Cookies and Intrusion Prevention Services (IPS). Connection limiting provides a means of throttling connections through the SonicWALL using Access Rules as a classifier, and declaring the maximum percentage of the total available connection cache that can be allocated to that class of traffic.
Coupled with IPS, this can be used to mitigate the spread of a certain class of malware as exemplified by Sasser, Blaster, and Nimda. These worms propagate by initiating connections to random addresses at atypically high rates. For example, each host infected with Nimda attempted 300 to 400 connections per second, Blaster sent 850 packets per second, and Sasser was capable of 5,120 attempts per second. Typical, non-malicious network traffic generally does not establish anywhere near these numbers, particularly when it is Trusted ->Untrusted traffic (that is, LAN->WAN). Malicious activity of this sort can consume all available connection-cache resources in a matter of seconds, particularly on smaller appliances.
In addition to mitigating the propagation of worms and viruses, Connection limiting can be used to alleviate other types of connection-cache resource consumption issues, such as those posed by uncompromised internal hosts running peer-to-peer software (assuming IPS is configured to allow these services), or internal or external hosts using packet generators or scanning tools.
Finally, connection limiting can be used to protect publicly available servers (for example, Web servers) by limiting the number of legitimate ingress connections permitted to the server (that is, to protect the server against the Slashdot-effect). This is different from SYN flood protection which attempts to detect and prevent partially-open or spoofed TCP connection. This will be most applicable for Untrusted traffic, but it can be applied to any zone traffic as needed.
Connection limiting is applied by defining a percentage of the total maximum allowable connections that may be allocated to a particular type of traffic. The above figures show the default LAN ->WAN setting, where all available resources may be allocated to LAN->WAN (any source, any destination, any service) traffic.
More specific rules can be constructed; for example, to limit the percentage of connections that can be consumed by a certain type of traffic (for example, FTP traffic to any destination on the WAN), or to prioritize important traffic (for example, HTTPS traffic to a critical server) by allowing 100% to that class of traffic, and limiting general traffic to a smaller percentage (minimum allowable value is 1%).
