Current Filter: Network>>>>>Feature>

   

Border Gateway Protocol, BGP for short, is the industry standard exterior gateway protocol that networks throughout the world use to communicate with each other. The Internet is essentially a giant web of interconnections between multiple Service Providers, some big, some small, and without them the consumers at home or at work would never have a complete view of the internet. BGP stores all of these destinations in a table and then uses it to route data correctly. Think of it like the GPS system in a car. BGP is the method used to store the destinations and all the available routes between them, and using this, it calculates the best route based on its own complex algorithm.

The average corporate user generally doesn't need or want to know how the Internet backbone works, that is until it goes wrong. Generally speaking, SMEs are very content and satisfied with a single service provider solution, as almost every provider can supply resilience in one way or another. Events like 512k day should not be direct concerns for the corporate user as it was a freak occurrence, and something completely beyond their control. For example, the last huge worldwide network incident was the world's biggest DDoS attack against Spamhaus which resulted in congestion across the globe.

Corporate users may be concerned about what technical information they need to know, how to manage it and how to do so without requiring too much in depth understanding. Essentially, every SME and Enterprise in the world should always completely understand how resilient their network is, and architecturally, how is it designed and built. Have you ever thoroughly tested your failover and resilience? From my vast experience there have been many times when, even with days and days of testing, there is one tiny part which is missed - and which can ultimately bring the entire platform to a halt.

512K day happened because the hardware memory limits used to store the BGP routes were exceeded. Many people have asked why engineers were not prepared for this apparent inevitability. However, even though every Network Engineer in the world had been watching for this day, the outages occured a lot sooner than expected.

There had been a lot of work in preparing seamless migrations carried out by Engineers throughout the world, working to a date much further in the future. Unfortunately, the steady growth which established the belief that there was at least 2-3 months of capacity left was undermined when someone, somewhere in the world, accidently made an error, releasing multiple routes.

I am often asked a very obvious question, namely "how can we make sure this does not happen again?" No matter how good your documentation is, or how fantastic your staff are, if your plans are not set in stone and tested on a regular basis then they are exposed to failure. As we all know, businesses evolve and plans change.

My response to this question is as follows: ask yourself these four key questions, and rate your answers between 1 (low) and 5 (high):

1. How thorough is your Disaster Recovery (DR) plan, and are you confident it works?
2. How confident are you in your provider's DR plan, and when did they last test it?
3. How confident are you about your entire setup?
4. How much risk are you willing to accept?

If you score less than 12 out of 20, then you must carry out an entire review of your infrastructure, as a matter of urgency. It may well be painful, but it is definitely essential. Resilience is the critical factor in network and business continuity and without it your business will always be at risk. NC