Previously, I discussed how rerouting data center host infrastructure should be part of next-generation DDoS solutions. In this blog, I will discuss how link availability solutions should also play a part. Traditional DDoS solutions offer us a measure of protection against a number of things that can disrupt service to our applications or environment. This is good, but what do we do when our mitigation solutions are downstream from the problem? In other words, what do we do if our service provider goes down either from a cyberattack or other event?
What if we had the capacity to clean the bandwidth provided by our service provider, but the service provider itself is down. How do we prepare for that eventuality? Admittedly, in first world nations with modern infrastructure, this is a less likely scenario. In third world nations with smaller carriers/ISPs and/or outdated infrastructure, it is more common. However, times are changing. The plethora of IoT devices deploying throughout the world makes this scenario more likely. While there is no silver bullet, there are several strategies to help mitigate this risk.
Is Border Gateway Protocol the Right Solution?
Most companies who consider a secondary provider for internet services have been setting up Border Gateway Protocol (BGP) as the service mechanism. While this can work, it may not be the right choice. BGP is a rigid protocol that takes a reasonable skill level to configure and maintain. It can often introduce complexity and some idiosyncrasies that can cause their own problems—not to mention it tends to be an either-or protocol. You cannot set all traffic to take the best route at all times. It has thresholds and not considered a load balancing protocol. All traffic configured to move in a certain route will move that way until certain thresholds are met and will only switch back once those thresholds/parameters change again. It can also introduce its own problems, including flapping, table size limitations, or cost overruns when it has been used to eliminate pay for usage links.
Any solution in this space needs to solve both the technical and economic issues associated with link availability. The technical issues are broken into two parts: people and technology. In other words, make it easy to use and configure; make it work for multiple use cases that include both inbound and outbound; and if possible eliminate the risk factors that can be associated with rigid solutions like link flapping and the associated downtime that can be caused via re-convergence. The second problem is economic. Allow people to leverage their investments’ fully. If they pay for bandwidth they should be able to use it. Both links should be active (and load balanced if the customer wants). A common problem with BGP is that one link is fully leveraged, and therefore hits its maximum threshold, while the other link sits idle due to lack of flow control or load balancing.
For several years, organizations have looked for alternatives. The link load balancing and VXLAN marketplace have both been popular alternatives, especially as it relates to branch edge redundancy solutions. Most of these solutions have limitations with inbound network load balancing, resulting in curtailed adoption. In many data centers, especially cloud deployments, the usual flow of traffic involves out-of-network users from the outside initiating the traffic flow. Most link load balancing solutions and VXLAN solutions are very good at load balancing outbound traffic. The key reason for the technology adoption has been two-fold: the ability to reduce cost with WAN/internet providers and the ability to reduce complexity.
The reduction in cost is focused on two main areas:
- The ability to use less costly bandwidth (and traditionally less reliable) because the stability was compensated for by load balancing links dynamically
- The ability to use what we were paying for a buy only the required bandwidth
The reduction in complexity comes from the ease in configuration and simplicity of being able to buy link redundancy solutions as a service.
The unique value of this solution is that you can protect yourself from upstream service outages or upstream burst attacks that trip thresholds in your environment and cause the BGP environment to transition back and forth as failover parameters are met, essentially causing port flapping. The carrier may not experience an outage, but if someone can insert enough latency into the link on a regular basis it could cause a continual outage. Purpose-built link protection and load balancing solutions not only serve an economic purpose but also protect your organization from upstream cyberattacks.