What is the difference between static and dynamic load balancing?

Static load balancing involves distributing tasks or workload evenly among processing units based on predetermined criteria, typically done at the start of execution and not adjusted during runtime. It doesn't adapt to changes in workload or system conditions. In contrast, dynamic load balancing dynamically adjusts the distribution of tasks or workload among processing units based on real-time system conditions, such as changes in resource availability or workload patterns. Dynamic load balancing continuously monitors the system and redistributes tasks as needed to optimize performance and resource utilization.

Static load balancing provides initial workload assignment to resources based on predetermined factors like task size or resource capacity. While it offers simplicity and speed, it lacks adaptability to changing conditions during execution, such as workload fluctuations or resource failures. Moreover, static load balancing requires prior knowledge of workload and resource characteristics, which may not always be available or precise. This method may be suitable for scenarios with stable and predictable workloads and resources.

Static load balancing is simple to use and it does not need to communicate with nodes real time, So, it need less compute power and does not affect system performance but it's great for the situation if you have relatively stable incoming traffic.

Static and dynamic load balancing are two approaches used to distribute incoming network traffic across multiple servers or resources in a computing environment. Here's how they differ: * Static Load Balancing: * Static load balancing involves configuring fixed rules or algorithms to distribute incoming traffic across servers or resources. * Load distribution decisions are made based on predefined criteria such as round-robin, least connections, or geographic proximity. * Dynamic Load Balancing: * Dynamic load balancing adjusts load distribution in real-time based on current system conditions, traffic patterns, and resource availability. * Load distribution decisions are made dynamically using algorithms that consider

Dynamic load balancing, gives you flexiblity for unstable incoming traffics but it needs to communicate nodes at all the time to understand the load to distribute traffics and that means extra compute power and it may impact system performance, at some point it may end up with under utilized nodes.

OpenMP used for Parallel computing on CPU threads within single node and MPI used for distribute parallel computing on multiple nodes. One important point in OpenMP is calculation requirements. If you need dependent calculation within all threads, you must activate atomic calculation method with following line of code : #pragma omp atomic and it will significantly affect the compute performance. If you have independent calculations for each thread, you don't need to use atomic method and that will improve compute performance. Another key method is using threads is omp_set_num_threads(4) if you use a number which is higher than your CPU threads, there will be no issue to run the code, openMP will provide maximum amount of available threads.

While working on a personal project I had multiple EC2 instances, and configured an Application Load Balancer (ALB) to monitor the health and traffic load of each instance. Based on real-time analysis, the ALB dynamically redistributed incoming user traffic to the healthiest instances with the lowest load. This made sure there was efficient resource utilization and application performance even as the demand fluctuated. This is key in production software to make sure there are no issues with performance.