E-commerce doesn't forgive slowness
A shop is not a brochure site. Every product page runs uncacheable PHP, queries the database dozens of times, and personalizes the cart. WooCommerce and PrestaShop alike generate between 60 and 150 SQL queries per page. On shared hosting, that load is paid for in seconds of TTFB — and in lost sales.
TTFB, the metric that decides everything
Time To First Byte sums up the health of your stack: CPU for PHP, IOPS for the database, network for transport. Our benchmarks, measured on customer shops after migration:
| Stack | Shared hosting | Titan VPS |
|---|---|---|
| WooCommerce product page TTFB | 1.4 s | 260 ms |
| Checkout funnel TTFB | 2.1 s | 340 ms |
| Sustained SQL queries / second | 1,800 | 24,000 |
Google factors page experience into its ranking, and every 100 ms shaved off TTFB translates into a measurable conversion improvement.
PHP-FPM: the number one tuning
The most common cause of a slow shop on a VPS isn't the VPS — it's a badly sized PHP-FPM pool. For a 4 vCore / 8 GB machine dedicated to WooCommerce:
pm = dynamic
pm.max_children = 24
pm.start_servers = 6
pm.min_spare_servers = 4
pm.max_spare_servers = 12
pm.max_requests = 500Rule of thumb: a PHP worker consumes 60 to 120 MB. 24 workers × 100 MB = 2.4 GB — leaving headroom for MySQL and the cache.
Absorbing spikes: sales, newsletters, influencers
An e-commerce spike isn't about averages: it's 30 times normal traffic for 3 hours. Two levers:
- Network headroom: with 10 Gbps, product images flow without saturating the link, even when 2,000 visitors arrive at once.
- Aggressive caching: catalog pages in full-page cache (Varnish or Nginx cache); only cart and checkout stay dynamic.
On our infrastructure, the shops that survive Black Friday intact share one trait: their peak CPU load never exceeds 50% of capacity in normal times.
The database on NVMe Gen4
The checkout funnel writes heavily: sessions, stock, orders, logs. On NVMe Gen4 at 1M IOPS, those writes never wait. On a saturated shared disk, the checkout is what slows down — the worst possible place.
Which plan for which shop?
Starting shop (under 5,000 visits/day): Titan. Large catalog, sustained traffic or multiple shops: Quantum. And size for the year's peak, not the average — a VPS resizes in minutes; a reputation for crashing during sales takes years to repair.