
We know who to blame
Blame the Cleaner. It's always that wee woman/man plugging in their 2000W vacuum cleaner next to the Router!
Amazon Web Services' Ashburn, Virginia, data center is giving the retailing and cloud computing giant headaches again, and Bezos can't blame the fickleness and severity of the weather this time around as he could back in late June when a hurricane-like line of thunderstorms ripped across the Ohio Valley and knocked out power to …
Say thank you if it is the cleaner.
I have seen what happens when an idiot plugs in an old "classic" welder in a UPS socket of a 6U AMC Symmetrix. It was ~14 years ago when only bank branches could afford such beasts. They forgot to tell the idiot who was installing the metal doors to the equipment room not to plug into it.
It was not a pretty sight...
To entrust your most precious data to a third party organisation that does not think twice about removing you should you have a minor infringement of the 3981 rules in the terms and conditions.
To entrust your most precious data to a third party that once having made up its mind will in no way discuss the reasons of the infringement.
To entrust your most precious data to a third party that changes the rules to suit it self.
You fools
Having your vital data on the cloud is not the issue. Having your vital on the cloud in one any only one place is the issue. If you put all your stuff in the same data center, then you're a) not doing cloud computing in the first place, and b) you haven't taken into account redundancy, which makes you no smarter than the idiot that has his vital information on a 20 year old hard drive he salvaged from a 386. When you're lazy about your failover design, cloud computing can't really help you.
"El Reg supposes that this is a fine line when it is your site that is essentially knocked out of commission by "performance issues" as cloudy systems go into super-slow mode and Amazon goes into cover-your-AWS mode."
Except all of the other EC2 DCs are working just fine?
Is the IT industry so prone to these absurd fads?
No doubt, internet-based on-line data storage has its uses. It is a good idea for one level of backup, at least, and I'm sure it's a godsend for accessing music on your phone, and all that sort of stuff. That anyone should use it as a primary data store, frankly, is beyond my belief. What's so hard about running a bloody file server, anyway?
To the person who has just sacked their server admin and signed away their data storage into this "cloud," I say, you sir are a bloody fool, possibly criminally negligent, and your auditors should drop a 15-ton weight on you. I also ask, how did you ever get into a position of such responsibility? A bullshitter promoted by bullshitters and buying from bullshitters? God help your company. I mean... companies: it seems that there are far, far too many of you about.
Because this isn't 1995. None of these sites are just serving files but assuming a file server will do the job it still does not account for redundancy and when your file server goes down you'll probably be knocked out longer as you wait for parts and rebuild the thing.
There is nothing wrong with aws if you have redundancy plans.
If and when that happens, at least I'm in charge, on the spot, and can be personally held responsible.
If this is "1995," then give me the 1995 version. 2012 seems to be utter madness. Use the cloud for backup. Be clever and switch to it, but primary availability? Well, I'm sorry to have to tell you I'm not offering you the IT director post.
Next!
The pattern of outages for the mega-cloud vendors (Amazon, Google, Microsoft, etc.) is disturbing. First there is the outages themselves. The June thunderstorm outage is an excellent case in point. No thunderstorm should take down a hardened data center. Why didn't they switch to diesel 30-45 minutes before the storms were obviously going to hit? A planned switch to diesel is far safer and more reliable than an automated emergency switch. If the planned switch fails, they have 30-45 minutes to take steps to get those diesels up and running or, at least, minimize the impact.
Second, they (after a week) still have not published a cause for last week's outage. I find that particularly curious. There is a definite pattern of poor communications from the vendors during and after outages. In the June incident, I noticed hours between updates and skimping on details. Enterprise-class customers will not suffer this treatment for long.
Third is that in many of these outages, the size of their environment seems to exacerbate the issues, causing such things as replication storms that back-up and cause even more problems. They claim to have availability zones, although some of these outages have affected multiples of them. Maybe they should further subdivide their environment and do a better job of isolating them. I believe they also share management functions across their environment. Maybe they should develop management zones aligned with their ideally smaller availability zones. At least offer this as an option, even if they charge a little more.
To me, this all points out the immaturity of the current cloud landscape and will drive enterprises more towards internal private clouds for the near- and mid-term until the vendors can get their acts together. It might also drive enterprises to the clouds being offered by the traditional outsourcing vendors who understand the enterprise market better.
Time for the cloud industry to do some soul-searching.