Luna Proxies For Browser Automation

Before choosing addresses, decide whether the job needs a browser at all. A real browser costs an order of magnitude more per page than an HTTP client, and that decision changes what you should buy.

Browser Or HTTP Client

If the data is present in the initial HTML, an HTTP client fetches it for a fraction of the cost and the proxy spend follows accordingly. A browser is necessary when scripts assemble the content, when interaction is required to reach it, or when the page checks for browser behaviour. Answering this first usually changes the volume estimate more than any proxy setting will.

What Automation Costs In Traffic

A browser loads everything: images, fonts, analytics, third-party scripts. Real consumption per page is commonly five to ten times the HTML alone, so a metered plan sized from page counts will be wrong unless that multiplier is included. Blocking non-essential resources at the context level is the single largest saving available, and it also speeds the run up.

Concurrency And Where It Caps

Throughput scales with workers only while each worker has its own exit. Past that point the limit is per-address rather than per-machine, and adding workers raises the error rate instead of the page rate. The useful measure is pages completed per hour at a stable error rate, not workers launched.

Throughput Against Cost

Automation cost has two components that move in opposite directions. Higher concurrency raises pages per hour but also raises the error rate once per-address limits bite, and failed pages still consume traffic. The efficient operating point is the concurrency where completed pages per gigabyte is highest, not where raw page rate peaks. Measuring both together usually reveals that the fastest configuration is also one of the more expensive ones.