Trends How Traffic Shaping Cuts Out 2/3 More Spam Than Blocking Alone By Ken Simpson | 2 minute read No, I don’t mean the movie. I’m talking about data that demonstrates how effective traffic shaping is against spam traffic. The following graph summarizes about one week of data from an ISP with several million users, showing how much traffic was culled through the use of Colored red: Two highly accurate block lists (one is the Spamhaus Zen list, and the other is a leading commercial black list used in some of the largest carriers worldwide) Colored orange: MailChannels Traffic Control 4.0 Colored yellow: Cloudmark‘s Authority spam filter (a very accurate carrier-grade filter) What’s significant about this graph is how much extra is being rejected because of traffic shaping. The orange region of the graph (28.7% of overall transactions) represents message deliveries that did not take place because the connections were slowed down by our Traffic Control system. Why was so much traffic rejected by traffic shaping? Spammers are impatient to get their messages delivered, and therefore have very little tolerance for slow connections. Traffic Control works by slowing down connections from new and otherwise not-yet-trustworthy senders, most of whom are spammers. When the spammer encounter these slow connections, they give up and move on to spam someone else. Because Traffic Control is highly efficient in its handling of large numbers of slow moving connections (we solve the c10k problem using libevent, for the developer geeks out there), the overall load on this ISP’s system was reduced so much that they were able to handle 25M transactions a day on a single eight-way Pentium-4 with 8GB of RAM (that’s with filtering, BTW). So if your job involves processing a huge amount of email and you’re wondering how to keep things under control without spending lots of money on new machines, consider giving us a call some time.