In recent weeks WikiLeaks has been targeted by denial-of-service (DoS) attacks, had its hosting service shut down, been bounced off of Amazon hosting, had its funding through PayPal, MasterCard and other sources shut down, and its leader arrested on sexual assault-related charges. The fact that WikiLeaks remains stubbornly and defiantly online holds some lessons for other sites when it comes to resilience and survivability.
Contrast WikiLeaks with Tumblr or Comcast--both of which experienced severe, prolonged outages in the past week. Or, compare WikiLeaks virtual invulnerability with the frequent and frustrating overloads and outages experienced by Twitter. Although the underlying motivation might be different, all Web sites and Web-based services can learn a thing or two from WikiLeaks.
In a word (or four), it's called "single point of failure". You don't want any. In fact, if you're WikiLeaks, you want to build redundancy on your redundancy and be able to survive not just a single point of failure, but a virtual meltdown of cascading failures.
WikiLeaks has implemented an interwoven network of domains, hosts, servers, and DNS services to ensure that losing any one provider, or having a server or domain shut down by a DoS attack will not ultimately impact the availability of the site.
WikiLeaks expanded from using a single DNS provider, to using 14 different DNS providers. WikiLeaks is directing traffic to at least five different IP addresses, and with the help of the WikiLeaks community-at-large it has more than 350 mirror sites hosting its content as well.
Bottom line: It would take a catastrophe of global proportions to completely knock WikiLeaks offline. Meanwhile, Twitter grinds to a halt if Lindsay Lohan is seen drinking a White Russian with lunch.
Of course, not every site can afford to build a redundant and interconnected network of Web hosting, domains, and Web servers capable of withstanding all manner of attack. Tumblr is a startup with a small team, a smaller budget, and a 24/7 business that has to keep up with rapidly growing demand. As noted on the Tumblr blog, "Frankly, keeping up with growth has presented more work than our small team was prepared for."
Even if your organization can't build a network with limitless redundancy and virtually invulnerable survivability, the lesson for IT and Web admins is to identify and eliminate as many single points of failure as possible. No one service outage or server crash should take your site down. Consider the entire network path of your Web traffic and--to the extent that it is possible and affordable--implement redundant secondary systems to eliminate and single points of failure.