Least Connections Algorithm: Dynamic Load Balancing

10/24/2024 •

Overview – Least Connections Algorithm

The Least Connections algorithm is a dynamic and adaptive load balancing method that distributes client requests based on the current load of each server. Unlike static algorithms like Round Robin, Least Connections considers the number of active connections on each server, directing incoming requests to the server with the fewest active connections. This helps ensure a more balanced load across servers, especially in environments where request processing times vary significantly.

How Least Connections Algorithm Works

Initialization: The load balancer maintains a list of all available servers and tracks the number of active connections on each server.
Request Handling: For each incoming request, the load balancer checks the current number of active connections on each server.
Assignment: The request is assigned to the server with the least number of active connections.
Connection Management: As requests are completed and connections are closed, the load balancer updates its records to reflect the current number of active connections on each server.

Detailed Example – Least Connections Algorithm

Imagine a scenario with three servers: Server A, Server B, and Server C. At a given moment, the number of active connections on each server is as follows:

Server A: 5 active connections
Server B: 3 active connections
Server C: 7 active connections

Incoming requests will be distributed based on the number of active connections:

Request 1 -> Server B (least connections: 3)
Request 2 -> Server B (now with 4 connections)
Request 3 -> Server A (least connections after B: 5)
Request 4 -> Server B (now with 5 connections, matching Server A)
Request 5 -> Server A (least connections after matching B: 5)
Request 6 -> Server B (now with 6 connections)
Request 7 -> Server A (least connections after B: 6)
Request 8 -> Server C (least connections after matching A and B: 7)

This dynamic allocation ensures that each server handles a load proportional to its capacity to process connections at any given time.

Advantages

Dynamic Load Balancing: Least Connections is highly effective in environments where request processing times vary, as it dynamically balances the load based on current server conditions.
Efficient Resource Utilization: By directing requests to the least loaded server, this algorithm ensures more efficient utilization of server resources.
Reduced Latency: With fewer active connections, servers can handle requests more quickly, reducing response times and improving overall system performance.
Scalability: The algorithm adapts to changes in server load and can scale efficiently with increasing numbers of servers and connections.

Disadvantages

Overhead in Monitoring: Continuously tracking the number of active connections on each server introduces additional computational overhead and requires more sophisticated monitoring.
Potential Imbalances: In cases where connections vary greatly in processing time, servers may still experience imbalances. For example, a server with fewer but long-running connections may appear less loaded than a server with many short-lived connections.
Complexity: Implementing the Least Connections algorithm is more complex than simpler algorithms like Round Robin, requiring more detailed configuration and management.
Fault Tolerance: Similar to other algorithms, Least Connections requires additional mechanisms to handle server failures gracefully. If a server goes down, the load balancer needs to detect this quickly and redirect traffic to healthy servers.

Enhanced Versions

To address some of the disadvantages, variations like Weighted Least Connections are used. In Weighted Least Connections, servers are assigned weights based on their capacity, similar to Weighted Round Robin. Requests are distributed based on both the number of active connections and the server weights, ensuring more powerful servers handle more connections.

Example of Enhanced Least Connections – Weighted Least Connections

Example

Consider three servers with different capacities. We assign weights as follows:

Server A: Weight 5
Server B: Weight 3
Server C: Weight 1

The algorithm calculates an effective load by considering both the number of active connections and the server weight. Suppose the active connections are:

Server A: 10 active connections
Server B: 9 active connections
Server C: 3 active connections

Effective load calculation:

Server A: 10 / 5 = 2
Server B: 9 / 3 = 3
Server C: 3 / 1 = 3

The request will be directed to Server A, which has the lowest effective load (2).

Summary

The Least Connections algorithm is a dynamic and adaptive load balancing technique that allocates client requests based on the current number of active connections on each server. This ensures a more balanced and efficient distribution of the load, particularly in environments where request processing times vary. While it offers significant advantages in terms of dynamic load balancing and resource utilization, it also introduces complexity and overhead in monitoring and managing connections.

Enhanced versions like Weighted Least Connections further refine the algorithm by considering server capacities, ensuring that more powerful servers handle a proportionately larger share of the workload. This makes Least Connections a versatile and powerful load balancing solution for modern, heterogeneous server environments.

Useful Links

https://www.a10networks.com/products

https://sanchitgurukul.com/tutorials-cat

Tutorial

Step-by-Step How-To Guides at Sanchit Gurukul