Hosting, CDN and traffic peaks

Hosting is the server (or cloud) the store runs on. A CDN (content delivery network) is a network of servers around the world that serves part of the site (images, scripts, sometimes pages) from a point closer to the shopper. Together they determine whether the site survives a sale and whether pages open quickly.

In plain words

When Black Friday brings several times more shoppers, a page has to be assembled for each one: the server queries the database, builds the html, serves the images. If resources run short, pages slow down, and then errors appear. So the hosting question is a question of revenue on the best days of the year.

What a CDN gives you

web.dev (the Chrome team) describes a CDN as a distributed network of servers that reduces the load on the origin server, cuts costs and copes well with traffic peaks. It brings content closer to the user, speeds up connection setup and serves cached resources without hitting your server. Even dynamic pages can be cached for a few seconds during a peak to relieve the server (web.dev).

Practical tips from the same source: - Cache images, video and versioned libraries for a long time (from six months to a year). - Private responses (cart, account) must not end up in a shared cache. - The share of requests served from cache is called the cache hit ratio; a typical target is around 90% (same source). - A CDN also helps speed: it shortens time to first byte (TTFB), which affects LCP, one of the Core Web Vitals metrics (same source).

Speed as a requirement

web.dev considers a TTFB under 0.8 s good and over 1.8 s poor; this is a benchmark, not a Core Web Vitals metric, but it helps diagnose LCP (web.dev). For the Core Web Vitals thresholds themselves, see Core Web Vitals.

What happens when the site can't cope

• Shoppers see errors and slow carts and leave. In Baymard's research, site errors or crashes are named as the reason for abandoning an order in 17% of cases (a US survey, data on a page updated in September 2025, without the year of the survey itself; Baymard).
• Search suffers. In response to 5xx and 429 errors, Google temporarily slows crawling; URLs that consistently return errors drop out of the index over time, and after recovery the crawl rate gradually rises (Google).
• Attacks. In its Q3 2026 report, Cloudflare recorded 18.2 million network DDoS attacks, 208% more than last year; 96% ended in under 10 minutes, and 98% were weaker than 1 Gbps. The largest attacks (over 1 Tbps) became fewer (Cloudflare). Takeaway: short, small waves are more frequent; protection has to work automatically, because you can't react by hand in 10 minutes. The report gives no advice to store owners, it only describes the attacks.

There is no fresh open data on Ukrainian stores' traffic during sales, so prepare from your own statistics of past peaks.

How to prepare for a peak

1. Look at the last peak: the maximum visitors and orders per minute, and when errors appeared.
2. Load test on a copy of the site a week or two before the promotion. Test the heaviest pages: search, filters, cart, checkout.
3. Cache and CDN for static files and pages that are identical for everyone (homepage, categories).
4. Switch off what you don't need during the sale: heavy widgets, extra calls to external services.
5. Scaling: ask your host how quickly resources can be added (manually or automatically) and what it costs.
6. Monitoring and on-call: who gets the alert if the site goes down on a Friday evening.
7. Plan B: a backup and a known recovery procedure; see backups.
8. Integrations: the warehouse, payment system and delivery service must also withstand the peak. Nova Poshta or a payment provider timing out produces the same picture as weak hosting.

Types of hosting

The type of hosting (shared, VPS, dedicated, cloud, a platform's managed service) determines how many resources are guaranteed to be yours and how many you share with others. For a store with peaks, what matters is not the name but the guarantees: resources, scaling, backups, support, server location relative to your audience. Cost of ownership touches on this topic too.

Common mistakes

• Choosing hosting by price and not testing for peaks.
• Caching nothing and serving every page from the database.
• Caching the cart and account for everyone (one shopper's data would be seen by another).
• Testing only the homepage, not search and checkout.
• Launching ads at the peak without warning IT.

FAQ

Does an online store need a CDN? Almost always, for images and scripts. It reduces the load on the server and speeds up loading (web.dev).

What is TTFB and what should it be? Time to the first byte of the response; under 0.8 s is considered good (web.dev).

How do I prepare a store for Black Friday? Load test, set up cache and CDN, switch off unneeded widgets, check scaling and monitoring.

What should I do if the site goes down during a sale? First restore the cart and checkout, tell shoppers about the outage, then work out the cause from the logs.

Related terms

• Core Web Vitals: LCP, INP, CLS
• Technical site audit
• Online store backups
• Online store security: the minimum
• Bots and price scraping in an online store
• What it costs to own an online store: startup and monthly expenses

Sources

• web.dev, Content delivery networks (CDNs) — web.dev/articles/content-delivery-networks
• web.dev, Time to First Byte (TTFB) — web.dev/articles/ttfb
• Google Search Central, HTTP status codes, network and DNS errors — developers.google.com/search/docs/crawling-indexing/http-network-errors
• Cloudflare, DDoS Threat Report 2026 Q3, 29 October 2026 — blog.cloudflare.com/ddos-threat-report-2026-q3
• Baymard Institute, Cart abandonment rate statistics — baymard.com/lists/cart-abandonment-rate