Amazon's S3 outage: Is the cloud too complicated?

Amazon's S3 outage: Is the cloud too complicated?

Summary: Over the weekend Amazon's S3 storage service was down for an extended period and a bunch of Web 2.0 sites lost avatars, images and other items on their sites.

TOPICS: Browser, Amazon, Cloud, Outage

Over the weekend Amazon's S3 storage service was down for an extended period and a bunch of Web 2.0 sites lost avatars, images and other items on their sites. Since enterprises haven't totally jumped on the bandwagon Amazon's outage didn't have broader ramifications. But Amazon's latest outage--the second big one this year--will hamper dreams of enterprise class services for the masses.

After all, the dream for cloud computing is enterprise reliability for pennies. In this view, the cloud will just work, uptime will always be there and we'll tap into this architecture and always be tethered to the Web. Michael Krigsman gives Amazon props for transparency with its latest outage, but the larger issue is reliability and how much redundancy should we expect for a few pennies a gigabyte (Techmeme).

If Amazon can't democratize cloud computing and bring us a bunch of "9s" reliability who can?

Om Malik writes:

The outage shows that cloud computing still has a long road ahead when it comes to reliability. NASDAQ, Activision, Business Objects and Hasbro are some of the large companies using Amazon’s S3 Web Services. But even as cloud computing starts to gain traction with companies like these and most of our business and communication activities are shifting online, web services are still fragile, in part because we are still using technologies built for a much less strenuous web.

Om hits the mark. The problem: The Web is one big legacy system. And cloud computing relies on millions of connections and services. In other words, it's a troubleshooting nightmare when the cloud goes bust.

And like any company wrestling with legacy systems cloud computing vendors will dust off a tired playbook. The solutions will be the usual: Relegate legacy systems to plumbing and create more services and applications to keep infrastructure current. In other words, the cloud will likely become more of a rat's nest. What's scary about that prognosis is the cloud is already too complicated since it's built on creaky infrastructure.

Also see:

Topics: Browser, Amazon, Cloud, Outage

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • It's a matter of infrastructure - and transparency


    Your analysis is partially correct, but there is a second factor. Some cloud providers have taken a "you're better off not knowing" approach to the infrastructure their systems are based on. What server is your VM on? What disk is your data on? How many switch hops between VMs? Is your data even in the same data center as your VM? Don't ask!

    Unfortunately, lacking this information, it's difficult for users to take standard architectural steps to provide reliability. Since many early adopters were small businesses, this issue didn't come to the forefront immediately. The unfortunate outages, though, are bringing the issue to the forefront.

    However, this lack of transparency isn't true of all cloud computing systems. At 3tera, our AppLogic system provides answers to all the questions I mentioned above and more.
  • There are so many reasons...

    using a third party service provider is bad for your IT. Down time is only a small reason. Privacy is probably the biggest reason. With deep packet inspection, and service providers handing information over to the government without warrants and/or subpoenas, just to name two, your data is at risk big time. Why any company would choose to put their data at risk is beyond me - with one exception - the price of bandwidth.

    Hardware is cheap, there are plenty of free, very good, open source solutions. With the exception of bandwidth, it can't be any cheaper to let someone else manage your data.

    The cloud will prove good for one thing - rain. It's a fad now but it won't last.
    • Try encryption

      You can use TrueCrypt on the data you store on the Cloud.
    • Exactly that: a fad!

      I believe that cloud computing won't last long. Operating systems, desktops, enterprise servers will continue to thrive in spite of the cloud computing fad. Everything cheap ends up costing you more. You basically get your money's worth.
  • RE: Amazon's S3 outage: Is the cloud too complicated?

    A Limerick For Our Times

    Someday in the cloud we'll all compute

    Of that there can be no real dispute

    But when the cloud disappears

    Global teeth-gnashing and tears

    In vain do we cry "Reboot! Reboot!"
    • Ah! A poet amongst us!

      Nice verse. :-)
      P. Douglas
  • RE: Amazon's S3 outage: Is the cloud too complicated?

    I fully agree! The same thing counts for any web based business application. Why would a company ever want to put its most important data on third party servers?

    How could any one decide to choose online business applications when customer/marketing/project information is involved?

    Maybe the bright side of this story is that it might scare off the uninformed.
  • RE: Amazon's S3 outage: Is the cloud too complicated?

    We already rely on "third-party" hardware and software all over the web -- DNS and routers and switches and backbones and communication links of all sorts. The design was aimed at nuke-resistance and operational capability anyway... I'm anticipating that we'll redesign the cloud cover, making it less complex and more robust. Redundant/alternative components can provide overall highest reliability without centralizing points of failure, and costs are better too. Simpler, stronger, less expensive -- it is possible.
  • this is why local will never go away

    or, at least, why the cloud will never go full 3rd-party for large business. You have to have offline capability, and for that you need a local solution. When online apps/storage are down, you still have to work.

    Since most cloud apps are going on/offline anyway, eventually there should be a backup solution for when the cloud is down - re-routing those websites using the cloud to a local backup, for instance.

    Everyone wants us to go either completely local or total cloud, when it's a good balance between the two that makes the most sense.
  • RE: Amazon's S3 outage: Is the cloud too complicated?

    When I hear the word "web 2.0" I always stop reading.
    • Hmmmm....

      ...I always start chuckling...
  • Cloud Computing is simply immature

    Cloud Computing, is just like every other infrastructure utility and needs time to mature. Surely, there will be mistakes and users will experience down time. I don't know of any enterprise IT shop that hasn't experienced unscheduled down time. Also, SLAs are missing from the discussion. They're are the key to understanding the real business impact.

    We rely on infrastructure services such as electricity, water, roads, telecommunications, etc., everyday which are far more essential than cloud computing. We still experience failures in these mature services such as power outages, closed roads, etc.

    Economics will drive the future infrastructure utility services such as cloud computing. It's inevitable.
  • It's not meant to be reliable

    S3 lets small-time developers do big box office, but that scalability comes at a cost in reliability. Higher reliability means a smaller audience. Try to put services on more than one device if possible.
  • Message has been deleted.

    Duke E. Love
  • The most important thing to remember

    [i]Boy: Do not try and bend the spoon. That's impossible. Instead... only try to realize the truth.
    Neo: What truth?
    Boy: There is no spoon.[/i]

    There is no "cloud" -- that's simply an abstraction. Behind the abstraction are physical systems that operate under the same principles as every other system in every other data center everywhere. To assume "the cloud" will always be available is a bit naive. The technology is just not there yet for that scale of reliability.
  • RE: Amazon's S3 outage: Is the cloud too complicated?

    Whomever said that it proves anything is off base. The fact of the matter is all of the storage centers with the "Put all of the eggs in one basket" philosophy are very easily interrupted.

    The datacenter is dead, it was dead when it was conceived. There are aplenty of solutions out there such as dedicated P2P networking that allows companies to have a distributed network around the world.

    But until all of those jokes of CEO's understand that P2P is not for them to download their latest in music or porn, it will never be a good fit for the enterprise.
    The Admiral