Google Apps contracts promise no 'scheduled downtime' Google has updated the contracts for its Google Apps suite so that they no longer make allowances for scheduled maintenance, and that any downtime - no matter how small - will be counted and applied to the customer's agreement. Mountain View announced the change to its Google Apps service level aggreements (SLAs) with a Friday …
They can promise no scheduled downtime *to the end user* because they have redundancy across multiple data centres. Anyone who knows anything about networking has looked at Google's operation purely because it's a brilliantly elegant way of taking reliability past that seemingly impossible 99.95% barrier.
For my operations, the only outages I'm afraid of are the ones that originate in the browser - since Google Docs is Java heavy, and most implementations of Java are leaky in the extreme, it's possible for my people to get major problems if they don't close and reopen the browser every few hours. If Mozilla, Opera, Safari, Chrome et al could lick that problem they'd have done humanity a real service.
Beer because it's Saturday out here in Australia.
What I think is impressive, isn't their promise of zero downtime, but the fact that they've added millions of users and added many new features without downtime.
What other service/IT system do you know of that could achieve this ?
Say what you want about the "Evil Google", but they've got some *real* smart cookies working for them.
It's easy. Anybody can do it, the problem is nobody wants to do it. When you friendly bank or supermarket goes offline for maintenance, ask yourself why they're completely clueless rather that considering it normal.
Update a percentage of servers at a time maybe? There's nothing complex about this. Google do it with more than one DC and all of a sudden you have an app that will never go down short of the world ending.
....*then*, optionally, comment. Thanks.
Hey just kidding. Go ahead and kneejerk away...even when the article actually conveys useful information essentially rendering the comment irrelevant. We need those kind of comments, for their simply distracting qualities, it's really quite useful.
and Happy New Year!
This is certainly moving towards how computing resources as a utility is supposed to work.
Now I'm thinking about how even well understood utilities can fail, and the general pattern of less frequent outages with much more widespread impact.
Will we inevitably one day see the Google equivalent of an un-pronouncable Icelandic volcano shutting down more than just a couple of planes? What's the plan then?
Surely this kind of offer is only interesting to those who already have contracts with several ISPs, carefully chosen so that they use different sets of cables running into the office (which of course has a whole-office UPS for when some tired JCB driver decides to plug your mains supply into your internet pipe.
Such people exist, but probably not in large numbers. I suspect this is mostly about bragging rights.
A promise of 3 9s availability - 99.9% uptime - is a number. What does it mean. Your old tip and ring telephone system operates at 5-6 9s! 100 to 1000 times better! better than 3 min/yr to .3min/yr downtime! The downtime of telephony is also pareto - this is the bad news! A few systems will have long downtime, e.g., an incident lasting a few hours while the remaining systems have absolutely no downtime. I don't know what google looks like. Based on outages you read about there are a lot of customers affected for a long time to give the lousy 3 9s. So one might assess that 3 9s is pretty bad!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
