Splitting and balancing the workload

As SREs, our first order of the day is to ensure we split and balance the workload among the shared resources and backing services. Partitioning the load impacts the reliability as it isolates failures to small sections of the total population. Balancing the load also improves reliability by assigning less loaded resources to attend to a new request, thus decreasing latency. We will understand each concept next.

Splitting

We split the workload based on the consumer profiles as a general practice. We can employ many criteria items to divide incoming requests, but some are frequent in the designs. If the solution is a multi-tenant product, splitting the load by the tenant is a must-have requirement. An independent resources group will serve each tenant. A second example is a global deployment where users are present in many countries. Having the load partitioned by geographical regions is a good practice. Multiple application instances must...