Cutting infrastructure load by 90%

Context

myITProcess is a Kaseya platform used by managed service providers to assess and grow their clients’ IT environments. Its data layer relied on synchronization jobs that had been scaled out across ten Kubernetes pods to keep up with demand on shared backend endpoints.

The problem

Most of that compute was idle or repeating work it had already done. The shared endpoints were saturated by sync traffic, and onboarding a new tenant meant adding more pods. Capacity planning was reactive, and cost grew with usage rather than with value delivered.

Approach

• Profiled the existing system to find where load was actually being generated, separating genuinely independent sync calls from duplicated ones.
• Re-architected the synchronization mechanism into a single, resource-efficient pipeline with deterministic batching.
• Cut redundant work on healthy datasets so that the system only pulled what had actually changed.
• Re-sized the pod topology against the new workload profile.

Outcomes

Kubernetes pods

10 → 2

Endpoint load

-90%

Stack