A CDN is designed to circumvent network congestion and protect against service interruption.
CDN benefits – reliability & redundancy
One of the essential characteristics of a CDN is its ability to keep website content online in the face of common network problems, including hardware failures and network congestion. By load-balancing Internet traffic, using intelligent failover, and maintaining servers across many data centers, a CDN is designed to circumvent network congestion and be resilient against service interruption.
What is load balancing? How does a CDN load balance traffic?
The purpose of a load balancer is to distribute network traffic equally across several servers. Load balancing can be either hardware or software-based. A CDN uses load balancing in a data center to distribute incoming requests across the available server pool to ensure that traffic spikes are handled most efficiently. By efficiently using available resources, load balancing can increase processing speeds and effectively utilize server capacity. Properly load-balancing incoming traffic is a crucial component in mitigating spikes in traffic that occur during atypical Internet activity, such as when a website is experiencing an unusually high number of visitors or during a distributed denial-of-service attack.
A CDN also uses load balancing to make changes quickly and efficiently when the availability of server resources fluctuates up or down. If a server fails and failover occurs, a load balancer will redirect the traffic allocated for the failed server and distribute it proportionally across the remaining servers. This provides resiliency and reliability by increasing the likelihood that hardware failures will not disrupt traffic flow. When a new server comes online in the data center, a load balancer will proportionately remove load from other servers and increase the utilization of the new hardware. Software-based load balancing services allow a CDN to scale load-balancing capacity quickly without the bottlenecks present when using physical load-balancing hardware.
What is failover? How does a CDN failover between servers?
In computer systems requiring high reliability and near continuous availability, failover prevents traffic from being lost when a server is unavailable. When a server goes down, traffic must be rerouted to a still functional server. By automatically offloading tasks to a standby system or another machine with available capacity, intelligent failover can prevent service disruption to users.
How does a CDN reliably serve content across the Internet?
A CDN is like a GPS coupled with express toll roads; a CDN will be able to find the optimal path to reach a distant location and will be able to use its network to find the optimal route to get there as quickly as possible.
When a user loads content from an Internet property such as a webpage or web application, a series of connections are made to reach the location at which the content is served. Network traffic is metaphorically considered a road and highway system; smaller surface streets move local traffic around the same area, and interstate highways transfer traffic into different states. When something goes wrong, like a tanker truck blocks all lanes of an interstate highway on the primary path into a different state, traffic must find another path around. To reach its final destination, traffic must move across different networks, like a highway crossing different states. If a blockage exists in a particular network, the traffic must be redirected down a different pathway. This process can be time-consuming and inefficient.
Let’s say a user in San Francisco is loading a website in Los Angeles. The connection makes many steps, but in this example, one of the most critical steps is where the network signal passes through a telecommunications provider based in San Jose toward the final destination. When a network engineer accidentally pours coffee on routing equipment in San Jose, the provider goes offline, breaking the connection (stranger things have happened). Users can no longer load their Internet content when this occurs unless the networked traffic is rerouted to accommodate the new network landscape. The user’s request must go through a different telecom provider if it ever wants to arrive in Los Angeles.
Now that the traffic can no longer pass through the intended network, it must step into an entirely different network maintained by a different organization. This process of renegotiation and switching networks may occur multiple times in a network request, and instances like this can add latency. They may push the traffic onto a congested pathway, delaying it. A CDN of sufficient size will typically control its network connections by placing servers in Internet exchange points (IXPs) and other strategic locations. These optimized network schemas allow CDN providers to optimize the route and reduce latency.
How can a CDN use an Anycast network to increase reliability?
Some CDNs will use an Anycast routing method to transfer Internet traffic to specific available data centers. This ensures improved response time and prevents any data center from becoming overwhelmed with traffic in the event of extraordinary demand, such as during a DDoS attack.
With Anycast, multiple machines can share the same IP address. When a request is sent to an Anycast IP address, routers will direct it to the machine on the closest network. If an entire data center fails or is otherwise incapacitated with heavy traffic, an Anycast network can respond to the outage somewhat similarly to how a load balancer transfers traffic across multiple servers inside a data center; the data is routed away from the failing location, and instead is routed towards another data center that is still online and functional.
DDoS attacks are currently one of the most substantial threats to the reliability of Internet properties. CDNs that use Anycast have additional flexibility in mitigating DDoS attacks. In most modern DDoS attacks, many compromise computers or “bots” are used to form what is known as a botnet. These compromised machines can generate so much Internet traffic that they can overwhelm a typical Unicast-connected machine. With an Anycast network, a portion of the botnet attack traffic can be distributed across multiple data centers, reducing the attack’s impact.
