pstoneman · HackerTrans

15 ปีที่แล้วdiscuss

The bit I really don't like is that it enforces systemd - and it sounds like it's part of it. I think one of the strengths of Unix/Linux is the architecture of many simple tools, each doing one job well, and one job only - which this sounds like it violates? Ick.

pstoneman

15 ปีที่แล้วdiscuss

I can't post on serverfault, since the question's been locked, so I'll put useful things to consider here:

* 100% SLA doesn't always mean 'It has to be up all the time'. Depending on the customer or the supplier, it can mean 'We'll aim to have it up all the time, but if it's not, we'll pay you compensation according to a predefined scale'. Clearly, in this case, you need to define quite firmly what 'up' and 'down' mean, how you measure them, how you time them, and how you decide what compensation to pay.

* DNS failover or load balancing is often nearly good enough. It won't get you instantaneous failover, since you'll need to have a finite (albeit small) TTL, and some client stub resolver libraries cache stuff anyway in violation of the TTL. But it's an easy step on the way

* If you want true 100% uptime, ultimately, you need a single IP (or range of IPs) which will be permanently reachable. That pretty much means the IPs need to come from one AS number - in other words, one ISP or one company.

* You can choose an ISP or company which has multiple internet connections, peers with a lot of people in multiple locations, and has a well-designed network such that you feel confident they won't go offline. Amazon may be a good example, but they've had several recent high profile failures!

* You could do it yourself - in which case, you'd need to become an ISP, get your own AS number, and set up peering arrangements with multiple suppliers in multiple locations. This can be very costly, and you still have to run a network and servers yourself in a reliable way

* You might be able to find a supplier who peers in multiple locations, and anycasts their protected IPs within their AS. That way, the same IP comes from multiple locations and should be reliable. Akamai might do something similar to this, I think.

* Ultimately, however you do it, you'll have a very difficult time making it impossible for it to fail. You're into the game of making it exponentially less and less likely that it'll fail, but you can't eliminate all risk. At the end of the day, your contract with your customer needs to define what happens if you should fail to reach 100% uptime. Is it breach of contract? Or do you need to pay a penalty fee? In either case, however you host it, you ideally need to make sure your suppliers compensation to you if they have a failure will cover the losses you incur.

pstoneman

15 ปีที่แล้วdiscuss

PC gamers who want such high specs might want to play them outside of their homes, when they go to LAN parties. This is fairly common amongst some gamers, and a portable powerhouse has some attractions for that use.

pstoneman

15 ปีที่แล้วdiscuss

I've seen banks do this before - and I think that some banks in some places use drop-downs rather than having you type characters. That's one reason I can see for the restricting the type of character so, so the drop-down is more sensible...

comments