4 ms·

If an hour's downtime would make you 'seriously consider' switching, I doubt you will find any service that meets your requirements. You're asking for 99.96% up

If an hour's downtime would make you 'seriously consider' switching, I doubt you will find any service that meets your requirements. You're asking for 99.96% uptime. I don't think any email provider will guarantee you that.

wwn_se 6y ago
Well gmail has 99.987% currently but only 99.9% is promised according to SLA. https://support.google.com/googlecloud/answer/6056635?hl=en#zippy=%2Chow-reliable-are-g-suite-and-google-cloud-platform https://support.google.com/googlecloud/answer/6056635?hl=en#...
- another-dave 6y ago
  And Protonmail offer 99.95% according to theirs (as of 2018, latest info I could find) — https://protonmail.com/blog/protonmail-reliability-sla/ https://protonmail.com/blog/protonmail-reliability-sla/
  - pmlnr 6y ago
    99.95% is a high standard already: 365x0.0005x24 is 4.38 hours outage for a year. In our current world, 4.38 hours is certainly not enough time to, say, rebuild a DB from backup, if needed.
    - throwaway81523 6y ago
      Yes, you fail over to a hot mirror, or if you really have to rebuild because something went nuts, do it from a recent snapshot.
    - another-dave 6y ago
      Yeah, especially when they're saying it's calculated on a calendar month, so you can't even "draw down" on previous months with 100% uptime if you have a serious outage.
      - cf_ 6y ago
        Yes, but I assume the rationale here is that they will only have to refund that month and can reset the clock for the next month. So if they have serious downtime it won't cost them for the full year - just for that one month.
interfixus 6y ago
Noone can gurantee any specific uptime. Any projection for the future, however qualified and propable, is still only a projection. GMail could go down tomorrow, for 24 hours or forever. (Not, of course, that such optimism is in way warranted).
- pmlnr 6y ago
  Whoever downvoted the comment above start your own server, and we'll talk a decade later. The comment is correct; nobody can guarantee uptime. The can promise, there can be agreements, insurance, etc, but guaranteeing is basically impossible. Even if someone were to use a real mainframe cluster across the globe one still connects to them via international ISPs and those routes can go down as well.

wwn_se 6y ago

Well gmail has 99.987% currently but only 99.9% is promised according to SLA. https://support.google.com/googlecloud/answer/6056635?hl=en#zippy=%2Chow-reliable-are-g-suite-and-google-cloud-platform https://support.google.com/googlecloud/answer/6056635?hl=en#...

another-dave 6y ago
And Protonmail offer 99.95% according to theirs (as of 2018, latest info I could find) — https://protonmail.com/blog/protonmail-reliability-sla/ https://protonmail.com/blog/protonmail-reliability-sla/
- pmlnr 6y ago
  99.95% is a high standard already: 365x0.0005x24 is 4.38 hours outage for a year. In our current world, 4.38 hours is certainly not enough time to, say, rebuild a DB from backup, if needed.
  - throwaway81523 6y ago
    Yes, you fail over to a hot mirror, or if you really have to rebuild because something went nuts, do it from a recent snapshot.
  - another-dave 6y ago
    Yeah, especially when they're saying it's calculated on a calendar month, so you can't even "draw down" on previous months with 100% uptime if you have a serious outage.
    - cf_ 6y ago
      Yes, but I assume the rationale here is that they will only have to refund that month and can reset the clock for the next month. So if they have serious downtime it won't cost them for the full year - just for that one month.

interfixus 6y ago

Noone can gurantee any specific uptime. Any projection for the future, however qualified and propable, is still only a projection. GMail could go down tomorrow, for 24 hours or forever. (Not, of course, that such optimism is in way warranted).

pmlnr 6y ago
Whoever downvoted the comment above start your own server, and we'll talk a decade later. The comment is correct; nobody can guarantee uptime. The can promise, there can be agreements, insurance, etc, but guaranteeing is basically impossible. Even if someone were to use a real mainframe cluster across the globe one still connects to them via international ISPs and those routes can go down as well.