AWS offers an SLA of 99.9 availability, which it has usually exceeded each year. That means your server can’t be down more than ~8h per year to beat it. Your residential ISP (in a nearly optimal case) has a 15-30 min service period overnight every few weeks.
Hope your area gets less than ~3 hours of power outages per year or you’re going to be breaching your SLA before you even hit software.
Your ISP is kind of dogshit if it’s forcing 15-30m of downtime overnight every few weeks. And power outages are kind of a weird thing to focus on, you should be on a UPS anyway.
In any case, someone interested in self-hosting email very likely has a redundant connection anyway. I’m not even hosting my own email and I have 5gb/mo of cellular backup in dual-WAN, and enough battery capacity to run my entire stack for several hours.
Not to mention a generator to recharge them, if it comes down to that.
Like, I need you to understand that in the networking industry, 99.9% uptime is genuinely laughable. You should be able to hit that by accident. The gold standard is ‘five nines’, or 99.999% uptime, or less than 5 minutes of downtime a year.
8 hours of downtime a year? If a service I was managing had 8 hours of downtime a year I would be laughed out of my job lol.
Your ISP is kind of dogshit if it’s forcing 15-30m of downtime overnight every few weeks. And power outages are kind of a weird thing to focus on.
Point being that these are not “skill issues”. AWS’s actual uptime over the last decade was something like 5 or 6 9s, 99.9 is just their official SLA. From where many people live (shit ISP, brown outs, floods, tornadoes, etc…), they can’t even match that bare minimum. God forbid budget enters the equation (no money for 3-2-1 backup? oops everything is fried from a freak accident).
So yeah you could definitely do OK with a real budget, a quality server setup and enough hours during the week for firefighting. But that’s not really “self hosting”, you’re just making your homelab a $0 revenue small business. For the 95% of people who can’t do that, they wouldn’t get anywhere close to a cloud provider’s service.
I would actually disagree- it doesn’t take much budget at all, or even a quality server setup, to have a decent uptime. A consumer router with a sim card slot is possibly something you already have. If not, a cell modem can be as cheap as $30. You could stick your email server on a old shitty raspberry pi. A data sim is $6/mo. If all you’re running is a cable modem, a router, and a rpi, you don’t even need a big fancy UPS, you can just get a DC battery UPS for like $40. And all this is assuming you’re buying stuff new instead of used.
You don’t need a lot of budget, quality stuff, or even a ton of hours in the week for self hosting- once you get this stuff set up it should stay working other than the standard upgrades/maint your email server will need.
Everything past that, like setting things up so your mail server is reachable on two IP addresses, is just… skill.
Lol the commenter you replied to didnt expect a one of a kind person to reply.
Normal people don’t have a ginormous battery and a generator for when the power goes out.
Every ISP is dogshit too. If it doesn’t go down from incompetence, it’s their physicial infra being broken from weather or some other “natural event”.
Even then, I can’t justify paying their crazy rates for 5g backup year round just for it to kick in once or twice a year or a couple nights where I’m not awake anyways.
Every email server that sends mail should have a rety mechanism if it fails to deliver too, so you shouldn’t miss any mail as long as your server isn’t offline for too long.
Ofc you are allowed to need 99.99% uptime for your home server, just disagreeing that it’s a need for most of us (including me).
I mean, my use case is abnormal and generally has more beef behind it than most people would have, yes, but a simpler, cheaper version of what I have set up is kind of a no-brainer if you want to self-host.
e.g. I don’t think a simple cyberpower/APC ups on your home server is any kind of a weird ‘specialty’ thing, and it should definitely run your server for 2-3 hours during an outage for like $100-150 if you grab it on sale (which, you know, why wouldn’t you?) As for the generator, I don’t have that for my network stack, I have that for my fridge/deep freeze lol. It can just also recharge my UPS if it’s really that big of a deal.
As for cell backup, that’s definitely less a ‘common’ homeserver thing, but I’m only paying like $10/mo for my cell backup connection from tello for 5gb of 5G. Hardly breaking the bank, and honestly probably overkill, you could likely get away with their $6/mo 2gb plan. No complaints with it either, I use them for my regular cell plan too. if you were interesting in self-hosting your own email server and wanted better uptime than 99.9%, you probably don’t even need that if your ISP only sucks slightly instead of mostly, but it allows you to just not care about your ISP having extended downtime and potentially timing out any retry mechanisms.
If you have a regular data plan that you use if your internet goes down, you’re server would technically have an internet connection, but your services (like email) still wouldn’t work would it?
Do you have some type of setup that keeps that working on data?
I don’t know how buying the ISPs data addon works, but I’ve been skeptical that the swap over would keep anything online either (but getting a generic data plan surely cant work at all right?).
So this sort of a setup is called Dual-WAN, and yes, it allows it all to work. Basically, my router has two connections to the internet- a cable modem on one port, which connects to Spectrum, and a cell modem on another port, which has my sim card on it. Both provide internet access simultaneously, and at that point, internet is internet- it doesn’t matter if it’s over data or through cable, you’re part of the net. My router is then configured to reach the internet via what it decides is the ‘best’ internet. In my case, because my cell connection is metered, I have it configured so that it prefers the spectrum connection, and only falls back to the cell connection if the spectrum connection is losing traffic, and only for as long as that connection is losing traffic.
Note that a dedicated cell modem is not necessarily required- some routers have sim card slots themselves, for exactly this reason, and tend to make this sort of configuration very simple to do. I’m personally using a small computer running OPNSense, which is again, probably overkill for the average homelabber, but you don’t need something that complicated.
As a result, my server always has access to the internet, and should you configure your firewalls appropriately (remember, I don’t personally run my own email- I have a custom dns name I point towards tutamail), the internet will then always have access to your server. There’s some details here and there about IP address caching, dns resolution, and the like which have various solutions from DDNS to an external proxy/loadbalancer/etc, but those are more implementation details.
Yeah fair, kinda funny how implied tone can work when reading things. On my first read, for some reason I took the “skill issue” as more playful but on second read, who knows.
Skill issue. My server has better online times than CloudFlare or AWS.
AWS offers an SLA of 99.9 availability, which it has usually exceeded each year. That means your server can’t be down more than ~8h per year to beat it. Your residential ISP (in a nearly optimal case) has a 15-30 min service period overnight every few weeks.
Hope your area gets less than ~3 hours of power outages per year or you’re going to be breaching your SLA before you even hit software.
Your ISP is kind of dogshit if it’s forcing 15-30m of downtime overnight every few weeks. And power outages are kind of a weird thing to focus on, you should be on a UPS anyway.
In any case, someone interested in self-hosting email very likely has a redundant connection anyway. I’m not even hosting my own email and I have 5gb/mo of cellular backup in dual-WAN, and enough battery capacity to run my entire stack for several hours.
Not to mention a generator to recharge them, if it comes down to that.
Like, I need you to understand that in the networking industry, 99.9% uptime is genuinely laughable. You should be able to hit that by accident. The gold standard is ‘five nines’, or 99.999% uptime, or less than 5 minutes of downtime a year.
8 hours of downtime a year? If a service I was managing had 8 hours of downtime a year I would be laughed out of my job lol.
Point being that these are not “skill issues”. AWS’s actual uptime over the last decade was something like 5 or 6 9s, 99.9 is just their official SLA. From where many people live (shit ISP, brown outs, floods, tornadoes, etc…), they can’t even match that bare minimum. God forbid budget enters the equation (no money for 3-2-1 backup? oops everything is fried from a freak accident).
So yeah you could definitely do OK with a real budget, a quality server setup and enough hours during the week for firefighting. But that’s not really “self hosting”, you’re just making your homelab a $0 revenue small business. For the 95% of people who can’t do that, they wouldn’t get anywhere close to a cloud provider’s service.
I would actually disagree- it doesn’t take much budget at all, or even a quality server setup, to have a decent uptime. A consumer router with a sim card slot is possibly something you already have. If not, a cell modem can be as cheap as $30. You could stick your email server on a old shitty raspberry pi. A data sim is $6/mo. If all you’re running is a cable modem, a router, and a rpi, you don’t even need a big fancy UPS, you can just get a DC battery UPS for like $40. And all this is assuming you’re buying stuff new instead of used.
You don’t need a lot of budget, quality stuff, or even a ton of hours in the week for self hosting- once you get this stuff set up it should stay working other than the standard upgrades/maint your email server will need.
Everything past that, like setting things up so your mail server is reachable on two IP addresses, is just… skill.
Lol the commenter you replied to didnt expect a one of a kind person to reply.
Normal people don’t have a ginormous battery and a generator for when the power goes out.
Every ISP is dogshit too. If it doesn’t go down from incompetence, it’s their physicial infra being broken from weather or some other “natural event”.
Even then, I can’t justify paying their crazy rates for 5g backup year round just for it to kick in once or twice a year or a couple nights where I’m not awake anyways.
Every email server that sends mail should have a rety mechanism if it fails to deliver too, so you shouldn’t miss any mail as long as your server isn’t offline for too long.
Ofc you are allowed to need 99.99% uptime for your home server, just disagreeing that it’s a need for most of us (including me).
I mean, my use case is abnormal and generally has more beef behind it than most people would have, yes, but a simpler, cheaper version of what I have set up is kind of a no-brainer if you want to self-host.
e.g. I don’t think a simple cyberpower/APC ups on your home server is any kind of a weird ‘specialty’ thing, and it should definitely run your server for 2-3 hours during an outage for like $100-150 if you grab it on sale (which, you know, why wouldn’t you?) As for the generator, I don’t have that for my network stack, I have that for my fridge/deep freeze lol. It can just also recharge my UPS if it’s really that big of a deal.
As for cell backup, that’s definitely less a ‘common’ homeserver thing, but I’m only paying like $10/mo for my cell backup connection from tello for 5gb of 5G. Hardly breaking the bank, and honestly probably overkill, you could likely get away with their $6/mo 2gb plan. No complaints with it either, I use them for my regular cell plan too. if you were interesting in self-hosting your own email server and wanted better uptime than 99.9%, you probably don’t even need that if your ISP only sucks slightly instead of mostly, but it allows you to just not care about your ISP having extended downtime and potentially timing out any retry mechanisms.
If you have a regular data plan that you use if your internet goes down, you’re server would technically have an internet connection, but your services (like email) still wouldn’t work would it?
Do you have some type of setup that keeps that working on data?
I don’t know how buying the ISPs data addon works, but I’ve been skeptical that the swap over would keep anything online either (but getting a generic data plan surely cant work at all right?).
So this sort of a setup is called Dual-WAN, and yes, it allows it all to work. Basically, my router has two connections to the internet- a cable modem on one port, which connects to Spectrum, and a cell modem on another port, which has my sim card on it. Both provide internet access simultaneously, and at that point, internet is internet- it doesn’t matter if it’s over data or through cable, you’re part of the net. My router is then configured to reach the internet via what it decides is the ‘best’ internet. In my case, because my cell connection is metered, I have it configured so that it prefers the spectrum connection, and only falls back to the cell connection if the spectrum connection is losing traffic, and only for as long as that connection is losing traffic.
Note that a dedicated cell modem is not necessarily required- some routers have sim card slots themselves, for exactly this reason, and tend to make this sort of configuration very simple to do. I’m personally using a small computer running OPNSense, which is again, probably overkill for the average homelabber, but you don’t need something that complicated.
As a result, my server always has access to the internet, and should you configure your firewalls appropriately (remember, I don’t personally run my own email- I have a custom dns name I point towards tutamail), the internet will then always have access to your server. There’s some details here and there about IP address caching, dns resolution, and the like which have various solutions from DDNS to an external proxy/loadbalancer/etc, but those are more implementation details.
And a fraction of the users so hardly comparable.
And yours does, that’s why you can’t self host your mail?
I’m not the other person. Try not being so combative dude. Jeez.
I can’t see how their reply was combative if yours wasn’t in the first place. Coming out of nowhere to protect AWS’s honour or something?
I mean, the person they initially responded to was being a bit combative too.
Yeah fair, kinda funny how implied tone can work when reading things. On my first read, for some reason I took the “skill issue” as more playful but on second read, who knows.
Playful banter or a deadly duel of words?
I know, but you engaged in this conversation to state your point and I am trying to understand how it relates.