Batching divides a large workload into smaller units that can be processed sequentially rather than all at once. It helps when a single pass exceeds available processing capacity or would hold too much data in memory. section 8.4.5 section 8.4.6
Generating invoices in manageable groups
Suppose an application generates thousands of monthly invoices, while the server can handle only 200 invoice generations at a time. Guava's Lists.partition divides customer IDs into fixed-size groups, and the application processes each group sequentially. section 8.4.5
List<Long> customerIds = loadCustomerIds();
List<List<Long>> batches = Lists #1
.partition(customerIds, 200); #1
for (List<Long> batch : batches) {
generateInvoices(batch);
}
The listing shows a complete batching flow: load the workload, partition it into groups of 200, and process each group. section 8.4.5
Apply batching with a few practical constraints:
- Choose a batch size that fits available processing capacity and memory instead of increasing infrastructure automatically. section 8.4.5 section 8.4.6
- Avoid batches that are too small when total execution time matters; efficient completion can return machines to low-power states sooner. section 8.4.5
- Combine batching with queues, asynchronous execution, or concurrency when workloads are unpredictable, user-triggered, or suitable for available CPU cores. section 8.4.5
- Use paginators, database cursors, or iterators when the main concern is limiting how much data remains in memory. section 8.4.6
