[Macworld]

Post your comments for Storms hit Amazon Cloud, downing Netflix, Instagram, others here

[cv]

This is a clear indicator that cloud computing *still* isn't ready for prime time.

Maybe in about ten years...

This post has been edited by cv: 02 July 2012 - 05:41 AM

[DocNo]

cv, on 02 July 2012 - 05:37 AM, said:

This is a clear indicator that cloud computing *still* isn't ready for prime time.

Because the sites affected didn't plan or pay for redundancy? If you notice Amazons other two data centers were still up. So poor architecture on the part of three sites is an indicator that all cloud computing "isn't ready"?

Fascinating.

Quote

Maybe in about ten years...

What's so special about ten years? Expecting the weather control network to come online?

In all seriousness, this is more a failure of facilities management on the part of Amazon. This the second time they have had an outage at that data center because of backup generators. If they are using a third party data center host, time to find a new one. If its managed in house time to put someone else in charge and get the core issues fixed. If I have a server or application in a commercial data center, loosing power - short of the building being destroyed - should not be a concern.

[Jerry2234]

What happened to unintrupted power supplies and emergency generators????? I worked in telecomm for 40 years and know there is no reason this happened except poor planning and too cheep to spend the $s necessary to prevent this from happening.

[soulatrium]

You just have to remember that the "cloud" is still tied to physical locations for wherever the servers are. This could be great when you're on the coast having a tropical storm, but your data is safe in VA. But of course it also cuts the other way; when there's a strong storm in VA, your data will be at risk to some extent if that's the only "cloud" location where you've stored it. But to me it sounds like a decent risk to take, VA certainly has fewer harsh weather occurrences than many other parts of the country.

So, any general statements about ready or not the "cloud" is are beside the point…the "cloud" will always have to have physical storage located somewhere, and it's really all about where that storage is and how it's managed.

[bastion]

soulatrium, on 02 July 2012 - 07:46 AM, said:

So, any general statements about ready or not the "cloud" is are beside the point…the "cloud" will always have to have physical storage located somewhere, and it's really all about where that storage is and how it's managed.

Which is why people that actually care about disaster preparedness have multiple redundant, geographically-dispersed storage systems.

But "the cloud" can never be "ready" because "the cloud" isn't an implementation; it's an idea. Any given provider of cloud-type services might be bettern prepared for some scenarios than others.

[KrasniOktabr]

Gently taps portable hard drive used for transporting files.

[elroth]

DocNo, on 02 July 2012 - 05:48 AM, said:

cv, on 02 July 2012 - 05:37 AM, said:

This is a clear indicator that cloud computing *still* isn't ready for prime time.

Because the sites affected didn't plan or pay for redundancy? If you notice Amazons other two data centers were still up. So poor architecture on the part of three sites is an indicator that all cloud computing "isn't ready"?

Fascinating.

Quote

Maybe in about ten years...

What's so special about ten years? Expecting the weather control network to come online?

In all seriousness, this is more a failure of facilities management on the part of Amazon. This the second time they have had an outage at that data center because of backup generators. If they are using a third party data center host, time to find a new one. If its managed in house time to put someone else in charge and get the core issues fixed. If I have a server or application in a commercial data center, loosing power - short of the building being destroyed - should not be a concern.

Every company knows about redundancy, every company tries to implement it, and sometimes they fail anyway. There is no guarantee of anything. "The cloud" is not failsafe - neither are nuclear reactors or anything else. This will happen again.

[Stewsburntmonkey]

elroth, on 02 July 2012 - 09:09 AM, said:

Every company knows about redundancy, every company tries to implement it, and sometimes they fail anyway. There is no guarantee of anything. "The cloud" is not failsafe - neither are nuclear reactors or anything else. This will happen again.

That's simply a truism though. Of course nothing is 100% reliable in technology (or anything else really). The question people have to ask themselves is how the cloud scores relative to the alternatives.

[MorrisTheCat]

I'm willing to bet that some companies located in Virginia in the region where these storms struck also lost power to their facilities and therefore lost access to their own internally housed systems. Bottom line is, this has little to do with cloud computing being unready as much as that power portages can affect anything, no matter where they're stored. If anything, proper "cloud computing" setups can make data access more available than just having it in one location, but it has to be implemented correctly. Evidently it wasn't in Amazon's case.

[JamieJamison]

cv, on 02 July 2012 - 05:37 AM, said:

This is a clear indicator that cloud computing *still* isn't ready for prime time.

Maybe in about ten years...

Nonsense. Your post is a clear indicator that you have no idea what cloud computing is all about. FYI Amazon maintains multiple data centers just as they maintain multiple distribution centers. Amazon has data centers in Virginia, Northern California and Oregon in the US and multiple data centers overseas. The point of this is to have geographical redundancy in case something happens to one data center. If NetFlix, Pinterest, Instagram or any of the other service providers had been running their own data centers in Virginia or were running out of someone else's co-location site then chances are that they would have lost power and suffered outages too. The reason why these companies don't have their own data centers is because building a resilient data center is very, very expensive. Any moron can purchase a lot of computers, stick them in a room and call it a data center, and many have. But in order to have something that's truly reliable you need to have multiple paths to the internet, not just multiple service providers but actual separate lines going out in different directions. This prevents you from losing connectivity due to what I like to call "The idiot with a backhoe" scenario, which happens more frequently than you would think. In addition to having a fairly complex network configuration you also need to have power management, with uninterruptible power supplies and backup generators. Engineering a reliable power system that can kick in if you lose grid power isn't cheap or easy. Then you need to have climate control for this data center, put several thousand servers, disk arrays and switches into a room and it gets very warm, very fast, oh, and that needs to be uninterruptible as well, so plan on purchasing some absolutely huge generators and uninterruptible power supplies to keep all of those tons of air conditioning running. Once you have that you also need to have a fire suppression system, so plan on laying out a whole bunch of money for an FM-200 system and then plan on spending money on the staff to keep all of this up and running.

If you're a smaller company you can't afford to provision a truly reliable data center, at best you can run your servers out of a co-location site. Oh, and if you want geographical diversity that means running your servers out of two co-location sites in separate areas. That's not cheap. Amazon Web Services, what you refer to as "the cloud" gives smaller companies a way to have an incredibly reliable infrastructure without having to shell out the large amounts of capital necessary to build that infrastructure. If I'm running my company's servers out of Amazon's us-east-1 region and decide that I need geographic diversity and want to run some of them out of their data centers in Oregon and Northern California it just takes a few clicks of a mouse and I'm in business. Amazon has some problems with the power management at us-east-1 that they need to fix, but outside of that their systems are ready for prime-time and incredibly reliable.

[JamieJamison]

MorrisTheCat, on 02 July 2012 - 10:04 AM, said:

I'm willing to bet that some companies located in Virginia in the region where these storms struck also lost power to their facilities and therefore lost access to their own internally housed systems. Bottom line is, this has little to do with cloud computing being unready as much as that power portages can affect anything, no matter where they're stored. If anything, proper "cloud computing" setups can make data access more available than just having it in one location, but it has to be implemented correctly. Evidently it wasn't in Amazon's case.

No, it is in Amazon's case. You're obviously unaware of the fact that Amazon has three AWS regions in the United States, us-east-1, in Northern Virginia, and two regions on the west coast located in Northern California and Oregon. The only region that was affected was us-east-1, Northern Virginia. Companies that were running out of the California and Oregon data centers didn't have any problems. Amazon is doing proper cloud computing, it's Pinterest, Instagram and Netflix, among others, who need to design their systems so that they're redundant across all of Amazon's data centers, which is very, very difficult to do.

[Stewsburntmonkey]

JamieJamison, on 02 July 2012 - 02:22 PM, said:

MorrisTheCat, on 02 July 2012 - 10:04 AM, said:

I'm willing to bet that some companies located in Virginia in the region where these storms struck also lost power to their facilities and therefore lost access to their own internally housed systems. Bottom line is, this has little to do with cloud computing being unready as much as that power portages can affect anything, no matter where they're stored. If anything, proper "cloud computing" setups can make data access more available than just having it in one location, but it has to be implemented correctly. Evidently it wasn't in Amazon's case.

No, it is in Amazon's case. You're obviously unaware of the fact that Amazon has three AWS regions in the United States, us-east-1, in Northern Virginia, and two regions on the west coast located in Northern California and Oregon. The only region that was affected was us-east-1, Northern Virginia. Companies that were running out of the California and Oregon data centers didn't have any problems. Amazon is doing proper cloud computing, it's Pinterest, Instagram and Netflix, among others, who need to design their systems so that they're redundant across all of Amazon's data centers, which is very, very difficult to do.

I don't think you can let Amazon off the hook. Their backup systems failed, which is not something that should happen. The US East region has consistently had reliability issues (it's the oldest one, so it makes some sense), so this isn't an isolated incident either. The way the Amazon EC2 system is set up makes is non-trivial to have cross region redundancy. As of a couple of months ago, when I last looked into it, there is no reliable way to dynamically load balance across regions for example.

Don't get me wrong, I love Amazon's web services, as they made my previous job running a small start up much, much easier. But there are some things they do need to work on.

This post has been edited by Stewsburntmonkey: 02 July 2012 - 03:28 PM

[RicD]

DocNo,
"...loosing power...". Oops, a common spelling error, it should read "..losing...". When spelling losing think "lost", each has one "o". The word loose means something is not tight.