Batching & Bulk Operations: Amortizing System Overhead

Every single network request, database query, or disk operation incurs a non-negotiable fixed overhead regardless of how tiny the payload is:

1. Network Layer Framing: Sending a 20-byte JSON message over HTTPS requires:
   • Ethernet Header (14 bytes)
   • IP Header (20 bytes)
   • TCP Header (20-32 bytes)
   • TLS Encryption Record Header (29 bytes)
   • Result: Over 80\% of the transmitted packet is pure network protocol overhead.
2. OS System Call Boundary Switches: Calling send() or write() triggers a CPU context switch from User Space to Kernel Space (sysenter / sysexit), invalidating CPU execution pipelines.
3. Database Transaction & Disk fsync Overhead: Every individual SQL INSERT requires starting a transaction, parsing SQL AST syntax, allocating locks, and executing a physical disk fsync() to flush the Write-Ahead Log (WAL) to NVMe storage.

The Power of Batching: By grouping 1,000 events into a single operation, you pay the network framing, syscall switch, and disk fsync cost exactly once, amortizing fixed costs and multiplying throughput by 10× - 100×.