Reference for Managing Website Bandwidth During Large-Scale Ad Campaigns

Quick Reference for Managing Website Bandwidth During Large-Scale Ad Campaigns

Below you’ll find a step-by-step guide for correctly handling the bandwidth of your Google Cloud instances when launching major advertising campaigns.

1. More traffic? Spin up more instances! What is an instance?
An instance is a virtual machine (VM) on Google Cloud—that is, a virtual server with dedicated CPU, memory, and storage for running your website.
When traffic spikes, we add instances so that load is spread evenly, ensuring fast response times and reliable service.

2. General rules (basic configuration)

• Default number of instances: 3 instances per country-specific site.

• Estimated capacity: With 3 F1 (Standard) instances, each handles ~10 RPS (requests per second), so 3 instances ≃ 30 RPS.

  • If each user makes 1 request every 5 s → 30 RPS × 5 s = 150 concurrent users

  • If each user makes 1 request every 10 s → 30 RPS × 10 s = 300 concurrent users

  This setup ensures:

  • Smooth browsing for about 150 concurrent users (and many more spread throughout the day).

  • Controlled hosting costs suited to normal site usage.

3. Specific rules for Ad campaigns
When launching a campaign, we may temporarily adjust the number of instances. Before scaling, we need:

1. Type of campaign: banner, newsletter, push notification, SMS, etc.

   • Newsletters generate gradual traffic (low risk).

   • App push notifications create very intense, simultaneous peaks (high risk).

2. Exact send time: confirmed date and start time.

3. Landing page URL: link to the campaign’s destination page.

With those details, we can scale resources only when needed, avoiding prolonged high-cost configurations.

4. Cost implications
Our goal is to keep hosting costs as low as possible by carefully managing traffic spikes.

• Concentrated peaks (e.g. 10,000 users in 1–2 minutes) drive significant cost increases.

• Distributed peaks (e.g. 10,000 users over 2–3 hours) are easier and cheaper to handle.

5. Reference table: Instances vs. Concurrent Users
Based on our tests (e.g. Affidea sites), here’s a quick lookup: