Skip to main content
Product Description Fault Tolerant Load Balancer
Last update:

Product Description Fault Tolerant Load Balancer

A fail-safe load balancer is a reverse proxy that distributes Internet traffic between different Selectel services located in different regions and availability zones:

Balancer can be used to build fault-tolerant services booked across multiple regions and hybrid infrastructures based on dedicated servers, cloud servers and VMware-based public cloud. The balancer also provides DDoS protection for all inbound traffic at the L3 and L4 levels.

Working principle⁠

To connect the balancer, you connect the infrastructure to a private network via Selectel global router, and connect the balancer to the same network. We provide you with an external IP address of the balancer to receive traffic, the address is protected from DDoS attacks and announced to the Internet via BGP Anycast protocol. Traffic arriving at the balancer is distributed in the infrastructure behind it according to specified rules — balancing algorithm, protocol, IP addresses and ports of target servers.

The balancer is booked in two independent regions, Moscow and St. Petersburg, with a pair of High-Avaliability balancers operating in both regions. If a router or balancer in one region becomes unavailable, traffic is automatically switched to another region.

Selectel is responsible for configuring the balancer, providing fault tolerance and monitoring.

Balancing levels⁠

According to the OSI model, balancing is possible at layers:

  • L4 over TCP protocol (transport layer) — the balancer redirects requests to servers according to specified algorithms;
  • L7 over HTTP (application layer) — the balancer analyzes the contents of the request and chooses which server to send the request to.

Balancing algorithms⁠

Different balancing algorithms are available at both balancing levels (L4, L7):

  • Round Robin is a round robin algorithm in which requests are passed to each server in turn;
  • Weighted Round Robin is a weighted round robin algorithm in which each server is assigned a weighting factor corresponding to its performance and power. More requests are transferred to servers with higher ratios;
  • Source IP hash is an algorithm that, depending on the HTTP header or IP address, selects a preferred server to receive the request;
  • Least Connections — an algorithm that sends the request to the least loaded server.

Comparison of balancing levels⁠

L4L7
Traffic TypeAnyHTTP
Public PortsAny443 (with TLS/SSL)
80 (without TLS/SSL)
ProtocolTCPHTTP
Support for SSL Offloading✗✓
Support for Let's Encrypt certificates✗✓
Redirect from HTTP to HTTPS✗✓ (with SSL Offloading active)
Support for HTTP/2✗✓ (with SSL Offloading active)
Forwarding of source IP addresses ✓ ✗ ✓ (with SSL Offloading active) Forwarding of IPsource addresses✓ (via proxy protocol, if enabled on backend)✓ (via standard X-Real-IP header)
Support active/standby✗✗
Session support✗✓ (sticky sessions)

Restrictions⁠

The product has no restrictions on the number of target servers and the amount of traffic to be transmitted.