The Offload Break-Even

Ada re-derives the offload break-even — time and battery for the transcode, local vs surrogate

ADA · CALCULATION AUDIT

Foundations · optional mathematics

The Offload Break-Even

The chapter transcodes a 4K clip two ways: locally it takes 320 s and drains 248.9 mAh, while offloading the three legs to a nearby laptop takes 149.3 s and just 18.7 mAh. Offloading only wins if transfer plus remote compute beats the local run. This audit computes the offload break-even to the second, showing it saves 53% of the time but 92% of the battery.

Companion to the chapter Edge Cyber-Foraging and Caching — every number here comes from that chapter.

Ada's Calculation Audit: the offload break-even — time and battery for the 4K transcode, local vs surrogate, ~4 minutes

Offloading only wins if transfer time plus remote compute beats the local run. Here is the break-even for the chapter's transcode, computed to the second.

The working

1. Time is size over rate; battery is current over time.

t = size ÷ rate    charge (mAh) = I × t ÷ 3600

2. Run it locally. 3840 MB ÷ 12 MB/s = 320 s. Battery = 2800 mA × 320 s = 896,000 mA·s ÷ 3600 = 248.9 mAh.

3. Offload the three legs. Upload 3840 ÷ 50 = 76.8 s; remote transcode 3840 ÷ 180 = 21.3 s; download 2560 ÷ 50 = 51.2 s; total = 149.3 s. Battery = 450 mA × 149.3 s = 67,200 mA·s ÷ 3600 = 18.7 mAh.

Path	Time arithmetic	Time	Battery
Local	3840 ÷ 12	320 s	248.9 mAh
Offload: upload	3840 ÷ 50	76.8 s	—
Offload: remote	3840 ÷ 180	21.3 s	—
Offload: download	2560 ÷ 50	51.2 s	—
Offload: total	76.8 + 21.3 + 51.2	149.3 s	18.7 mAh
Saving	(320−149.3)/320; (248.9−18.7)/248.9	53.3%	92.5%

What the audit buys you: the phone's radio draws 450 mA but for only 149 s, while its CPU draws 2800 mA for 320 s — so offloading wins on battery (92%) by a far wider margin than on time (53%), because it is the surrogate, not the phone, that pays the compute energy.

Every number above is taken from the chapter's own cyber-foraging offload example and re-derived step by step.