One Year of Hosting in Palitana: What We Now Know

Here is what went on in our first year. We look at the numbers, the down times, and what we plan to do next. It is a plain look at the work.

We have run Palitana for one year. Here is what we found. What went well, what did not, and what comes next.

The numbers.

We kept quiet at launch. Now we can share.

Latency. Half of all visits in India take 25 milliseconds. Visitors in India load pages in under 25 ms. Going faster depends on their own provider. That is not on us.
Uptime. See /status for the live feed. It is the only source.
Workload mix. WordPress, ERPNext, and one-off apps. See / for how we deploy.
Support. India hours, plus a small shift in the US for Europe. WhatsApp first. Email when WhatsApp is down.

Three events shaped how we manage risk. Each one has a write-up. /status has the live feed.

The generator switch broke. The battery backup held for 30 minutes. The generator did not start. We now test the switch four times a year.
A cache key bug. Static files shared one key across URLs. The files wrote over each other. We now use the full URL.
The database pool was too small. We set it to 10. We needed 40. We now load-test four times a year.

Live records sit at /security/incidents.json.

Latency. Guests in India load pages faster. Most of our users live in India. That matters.
WhatsApp support. A real engineer reads your message in the same shift. No portal. No email queue. We do not publish a response time here. /status and /contact have the real numbers.
Rupee pricing. The plan price is the bill price. No currency surprises. We can send a goods and services tax (GST) bill on request.

Write down what to do. Our guide said "transfer to generator after the battery bridge." It did not cover a broken switch. We now keep a 2-page guide for it. We drill it four times a year.
Watch everything. The cache bug took 4 hours to find. We had no screens to look at. We now watch the same gear. Alerts fire in 30 seconds.
Always plan ahead. Load tests on launch day are not enough. We run them four times a year.

A second region. Today we run just Palitana. Bangalore comes first next quarter.
Edge compute. We route calls at the edge of the content delivery network (CDN). We ship a small Node service for it. It is not a feature the CDN sells.
GPU plans. Some clients asked for containers with a graphics processing unit (GPU). Our top plan has the slots. We list the price on /pricing when it is ready. Message us on WhatsApp for early access.
Cheaper backups. Daily backups sit on S3 Standard. We will move them to S3 Infrequent Access. That saves money. It goes with the 7-day retention upgrade.