Client asks for 100% uptime • Hacker News Board

Re: Client asks for 100% uptime

#11

Post by rwmj » Fri, Sep 30, 2011, 12:10 PM UTC

Come on, this is possible.

First we're going to have to get the governments of the world together to agree to remove all nuclear weapons. Second would be getting that asteroid tracking and deflection system working. Quantum physics does unfortunately predict that the earth might flick out of existence with some small probability, but by distributing the website across the universe we can reduce this probability arbitrarily (and numbers approaching p=0.9999.. are the same as p=1). The client is going to need to budget for this.

Re: Client asks for 100% uptime

#12

Post by brador » Fri, Sep 30, 2011, 12:13 PM UTC

ck2 wrote: post #6

DNS round-robin with mirror servers that run 24/7.

Yup, this is what I'd do. Mirroring would be the big issue, but not that big a deal unless data was time critical.

Re: Client asks for 100% uptime

#13

Post by Toenex » Fri, Sep 30, 2011, 12:14 PM UTC

I would agree to this but charge them infinite money.

Re: Client asks for 100% uptime

#14

Post by _corbett » Fri, Sep 30, 2011, 12:15 PM UTC

I was at an Akamai presentation the other day in which the salesperson claimed "100% reliability" of their services "not 99%, not 99.9% but 100%".

Re: Client asks for 100% uptime

#15

Post by gaius » Fri, Sep 30, 2011, 12:22 PM UTC

joelhaasnoot wrote: post #5

eli wrote: post #4

Earlier quoted context omitted.

I've definitely seen hosting services do this. Sure, there's a 100% SLA, but if you actually read it, it says you get back the pro-rated monthly fee for the time it was down. So, in other words, you don't have to pay for it when it isn't working. Not much of an SLA.

Public transport does this too: in order to provide a robust, perfect implementation of a schedule, you need extra busses/trains/street cars (expensive capital goods) and extra manhours (generally most expensive part of the operation). Rather than invest too much money in making sure the schedule can be met, it's cheaper to pay fines when there are delays.

British Rail once declared they operated on a 66-minute hour for this reason.

Re: Client asks for 100% uptime

#16

Post by ck2 » Fri, Sep 30, 2011, 12:28 PM UTC

brador wrote: post #12

ck2 wrote: post #6

DNS round-robin with mirror servers that run 24/7.

Yup, this is what I'd do. Mirroring would be the big issue, but not that big a deal unless data was time critical.

Sites with large amounts of data do this in realtime all the time.

WordPress.com for example has five mirrors (maybe more by now).

Re: Client asks for 100% uptime

#17

Post by babebridou » Fri, Sep 30, 2011, 12:30 PM UTC

Would a heavyweight client with nothing but static data and no network at all reach 100% uptime, from a contractual point of view? Even a wrist watch does not exactly guarantee 100% uptime.

Re: Client asks for 100% uptime

#18

Post by Duff » Fri, Sep 30, 2011, 12:31 PM UTC

Their craziness doesn't matter. Usually crazy customers aren't rich. So if you build to their craziness, you'll lose the customer.

You need to build an appropriate infrastructure that will win the bid, figure out what you can achieve (99.9%/99.99% uptime) and build in enough overhead to cover your SLA penalties. Or negotiate a monitoring methodology that is in your favor. (ie. exclude planned maintenance windows, use a monitoring threshold/interval to allow you to address issues before triggering contract "downtime", exclude external provider issues, etc)

Re: Client asks for 100% uptime

#19

Post by viraptor » Fri, Sep 30, 2011, 12:37 PM UTC

ck2 wrote: post #6

DNS round-robin with mirror servers that run 24/7.

Various DNS caching architectures deployed by ISPs will sometimes strip additional addresses, or choose one for you themselves. That means if you're unlucky, that selected address will be down. Even if they do handle DNS correctly, hitting a dead server and failing over to the next one (possibly also dead) can take some time and may be interpreted as service not working.

Unless you control the whole path to the customer and his client application, DNS is not a solution for either HA or LB. (found out after trying to come up with something close to 100% for voip network - LB with 2 DNS entries on public internet will give you split close to 70-30)

Another question is how is the downtime interpreted? If you're halfway through some transaction / flow and the site you're talking to goes down, should your flow (shopping basket for example) be available on the failover site?

Re: Client asks for 100% uptime

#20

Post by sunchild » Fri, Sep 30, 2011, 12:44 PM UTC

100% uptime is not an operational requirement – it's contractual. A client that demands 100% uptime isn't being unreasonable; they're looking for a contract remedy (most likely a termination right) if/when the site goes down.

1. "Uptime" is defined in many, many ways. In the OP's article, it's the definition of uptime that seems unreasonable. Normally, the demarc points for the network segments and equipment being measured for uptime are entirely within the provider's control. In the OP article's update, the client clarified that 100% uptime only applies when hosting is cut over to the provider's site – something they are (theoretically) capable of controlling.

2. Remedies for failing the uptime requirement are different for nearly every agreement. Often SLA credits are the exclusive remedy. Sometimes the customer has a termination right (either express, or through the termination for cause provision). The remedy is probably more important than the uptime percentage.

You'd be surprised how many big name web apps offer 100% uptime as a matter of contract, knowing that it's a near-impossible operational goal. It's a matter of taking on the risk of your customer leaving you or claiming SLA credits, or whatever remedies you agree upon.