Introduction

In many business systems, core data is not static. Changes in business rules, financial parameters, or regulations can require recalculating large volumes of existing data.

These recalculations are typically:

• Computationally intensive
• Time-sensitive
• Infrequent, but critical

Traditional systems are not well suited for this pattern. They often rely on dedicated infrastructure that must remain available, even though it is only heavily used during limited periods.

Context: insurance and pension systems

This pattern is common in financial and insurance domains.

In this use case, a pension insurance system required recalculating all individual plans whenever new parameters were introduced for a given timeframe, such as indexation or profit conditions.

This meant:

• Processing large datasets
• Applying updated business rules consistently
• Ensuring correctness across all contracts

The challenge of occasional heavy workloads

These recalculations do not happen continuously. They are triggered only when:

• Financial parameters change
• New business rules are introduced
• Regulatory updates take effect

Traditional approaches typically rely on:

• Dedicated batch servers
• Fixed infrastructure sized for worst-case scenarios

This leads to:

• High infrastructure costs
• Low utilization outside recalculation periods
• Limited flexibility when workloads increase

In practice, systems are either:

• Overprovisioned (expensive most of the time), or
• Too slow during critical recalculation windows

Additionally, running large recalculations in parallel can put significant pressure on underlying systems such as databases, leading to contention, throttling, or failures.

The redesign approach with Taskurai

Instead of relying on fixed infrastructure, recalculation workloads are offloaded to Taskurai as asynchronous, durable tasks.

Key principles:

• Each recalculation unit is executed as a Taskurai task
• Workers process tasks in parallel
• Execution is fully asynchronous and durable
• Compute capacity scales automatically based on demand

At the same time, Taskurai allows controlling how work is distributed and executed, ensuring that downstream systems such as databases are not overwhelmed.

This shifts the model from:

👉 Fixed infrastructure sized for worst-case
to
👉 On-demand execution with controlled throughput

Handling large-scale recalculations

With Taskurai:

• Large datasets can be split into smaller tasks
• Tasks are executed in parallel across multiple workers
• Progress is tracked centrally
• Failures can be retried or resumed automatically

In addition, execution can be tuned to balance load:

• Control the number of concurrent workers
• Limit throughput towards specific systems
• Spread processing over time when needed

👉 This makes it possible to scale compute without overloading critical dependencies

During recalculation periods:

• Workers scale up to handle the workload

Outside of these periods:

• Workers scale down to zero

Operational visibility and control

Running large recalculations introduces operational risk, especially under time pressure.

Taskurai provides:

• Centralized tracking of all recalculation tasks
• Clear visibility into progress and completion
• Detailed logs and failure diagnostics
• The ability to retry or resume tasks without restarting the entire process

This improves:

• Operational control during recalculation periods
• Time to resolution when issues occur
• Confidence in the correctness of large-scale processing

Business impact

Reduced infrastructure costs

• No dedicated batch servers required
• No idle infrastructure between recalculation periods
• Lower operational and maintenance overhead

Faster processing with controlled load

• Parallel execution reduces total processing time
• Throughput can be tuned to avoid overloading databases and external systems
• Large datasets can be processed safely within required timeframes

Improved flexibility

• Easily adapt to new business rules or regulatory changes
• Control execution speed based on system capacity
• No dependency on fixed batch environments

Modernized architecture

• Transition away from legacy batch processing systems
• Easier to maintain and extend
• Better integration with modern applications and services

Summary

By using Taskurai to execute large-scale recalculations on demand:

• Heavy processing is handled efficiently
• Infrastructure costs are reduced
• Workloads scale dynamically with demand
• Downstream systems remain protected through controlled execution

All without relying on dedicated systems that are only used occasionally.