Fastly, a content distribution network (CDN), experienced an outage on June 9 that knocked out several high-traffic sites, including The Guardian and The New York Times newspapers, a number of British government sites, Reddit, and Amazon.com. The company attributed the outage to a bug in software Fastly was using to upgrade a client’s infrastructure. The company apologized for the problem in a blog post, saying, “This outage was broad and severe, and we're truly sorry for the impact to our customers and everyone who relies on them.”
As the outage occurred, web users trying to access affected sites received error messages like “Error 503 Service Unavailable.” The outage started just before 6 a.m. EDT. Fastly started implementing a fix and recovery in less than an hour. Still, the impact was significant, with Fastly suffering some reputational damage.
Difficult as this episode was for Fastly, it should serve as a reminder that outages and unanticipated configuration problems can happen to any provider of cloud services, backup servers, and CDNs. The outage also highlights how dependent businesses have become on third parties to deliver their online presences. Not even major brands like The New York Times and Amazon are immune from the effects of seemingly minor problems elsewhere on the Internet.
IT Central Station members are dealing with the same challenges. As they explain in reviews of products such as Cloudflare and Imperva Incapsula, reliable configuration and scalability are critical to success. For example, an information technology engineer at a tech vendor who uses Cloudflare shared that “it is easier to configure and develop documentation to see how we have configured firewalls. It's also more automated.” Their use case is more security oriented. This user cited the example of the risk of SQL injection, which Cloudflare can catch before it infects their infrastructure.
For Brian W., a VP/director of IT who uses Cloudflare, what matters is stability. Everyone wants to avoid an outage like the one that affected Fastly. He said, “Stability is great. I love their mechanisms, I love their GUI, I love their front end, I love their behind the scenes. Algorithms and everything, it all makes sense.”
A Cloudflare user who works as a network security engineer at a software company further spoke to this issue. He said “The stability is good. We haven't witnessed crashes, bugs, or glitches. It's a reliable product.” Raj M., president and CEO at Mekas Cloud Services, echoed this sentiment, commenting, “I haven't had any issues with stability [with Cloudflare]. We haven't heard anything from our clients. It's a stable solution.”
Users of Imperva Incapsula also weighed in on their solutions’ ability to keep infrastructure running reliably. As Ajay J., technical sales director at Revere Technologies, put it, “The product meets the needs of our business model and we can see the health of the architecture at a glance.” He was also pleased that the solution meets client compliance needs and offers auditing capabilities.
“The solution's most valuable aspect is that it is easy to configure,” said a solutions architect at a financial services firm. He added that “the solution keeps itself up to date itself and there's no customization that we need to do. It makes it extremely easy and cuts back on the amount of work required, and saves us on man-hours.”
Imperva Incapsula’s stability is what stood out for a vice president of global IT security at an insurance company. He noted, “The stability of the solution is very good. I've never heard of any outages that I'm aware of. It doesn't crash or freeze. There aren't bugs or glitches. It's reliable.”
As these real user reviews show, different CDN and cloud infrastructure solutions offer their own distinct advantages for stability and ease of use. The risk of a Fastly type outage is always a possibility, so users have to be diligent in their patching and upgrading processes. Getting things right and avoiding downtime requires a mix of people, processes and technology. The solution itself is neither the sole root of a problem, nor the lone barrier to unplanned outages.