This. Host says "scheduled maintenance, excluded." You say "no notice given." Who wins? The one who wrote the contract. I had a provider claim a 3-hour outage was "network upstream issue" and therefore excluded. My monitoring showed their whole ASN down. They still denied the credit. I moved.
Singapore problem: AWS, GCP, Azure all have SLAs with teeth but good luck getting money out of them when the issue is "regional." Last year I had an AZ issue in ap-southeast-1, two services affected. AWS credited me 10% for one, denied the other as "not simultaneously unavailable." Same incident, different products, different outcomes. Multi-cloud means I get to compare their excuses in real time.
Chen you are paying for five providers to avoid failure and getting five times the billing complexity. I am in Bangalore, I run four cheap servers like NUT. Total monthly: less than your cheapest single provider. No SLA on any of them. I accept downtime, my users are not paying per minute. Different math.
Same. Bangkok, four servers, mix of local Thai host and RackNerd. One is $12/year. The others are more but still under $20/mo total. Django app, mostly async tasks, can survive a node down for hours. I have backups to Wasabi. SLA is not in my vocabulary.
I used to chase SLAs. Spent more time arguing credits than the credits were worth. Switched to "acceptable downtime budget" model. For my actual revenue-generating stuff: two providers, both with defined compensation, one in EU one in US. For the rest: cheapest that works. The spreadsheet energy went into monitoring instead.
Following this thread. So if I understand: for a personal project, don't worry about SLA. For business, demand compensation defined and independent monitoring. Is that the summary?
Close. Add: read the exclusions list. If it's longer than the promise, the number is decoration. And verify if "availability" means control panel, network, or VM. Some hosts exclude everything but the website you use to open tickets.
Brazil perspective: local hosts advertise 99.99% but the compensation is "we'll investigate." International hosts with real SLAs often price in USD, which hurts when Real drops. I have one client on OVHcloud — https://www.ovhcloud.com — because they bill in BRL and have a real SLA. Not the best uptime, but at least I know what I'm getting.
Do they? Last I checked OVHcloud Brazil bills in BRL but the SLA is the same as EU: 10% per hour over threshold, max 50% monthly. Not terrible. I used them for a project in Portugal before moving to Hetzner. Support was slower but credits came through.
Quick mod note: thread's been excellent. Just confirming OVHcloud SLA discussion is fine since it's public policy, not internal detail. Carry on. Also @SuperNoob your summary is decent, read the first page for more nuance.
