Video: Is it time to throw away your servers?

Video: Is it time to throw away your servers?

Summary: Last November, I wrote a post with the title Amazon’s Jeff Bezos: Honey, I just shrunk the server hosting business. In that post, I did some simple math to show how a small business like the one I run with Doug Gold (Mass Events Labs, where we host two servers with a bigtime hosting service) can save some serious money if it can figure out how to move from a physical hosting service to an "elastic" one like Amazon's Elastic Compute Cloud (EC2).

SHARE:
TOPICS: Servers
6

Last November, I wrote a post with the title Amazon’s Jeff Bezos: Honey, I just shrunk the server hosting business. In that post, I did some simple math to show how a small business like the one I run with Doug Gold (Mass Events Labs, where we host two servers with a bigtime hosting service) can save some serious money if it can figure out how to move from a physical hosting service to an "elastic" one like Amazon's Elastic Compute Cloud (EC2). So, if a small business can save some serious money on two servers, can you imagine what a larger company with more servers can save. Unfortuantely, right now, it's not as simple as calling up Amazon and asking them to turn on two servers the way you'd do that with a hosting service. An EC2-based server gets launched through Amazon's web services APIs and, in the same way you can programmatically launch a server on the fly and provision it, you can de-commission it on the fly as well. Since EC2 is a pay as you go service, the advantage is that you only pay for the server capacity that you need. Unlike with a traditional hosting facility where, once you sign up for a couple of servers, you're probaby stuck with them for a year. So, given this dynamic server on demand model, EC2 represents another opportunity to save money than just the sheer cost. Anyway, I decided to do explain this concept through one of ZDNet's trademark whiteboard videos. Check it out.

Topic: Servers

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

Talkback

6 comments
Log in or register to join the discussion
  • cluster computing

    Great video. It lays out the numbers for why cluster and grid computing stand poised to "revolutionize" the web services business. Rather than dealing in commodity computers, we will begin to deal in commodity computing cycles (C3), where we are purchasing FLOPS, not boards.

    There are some hurtles in stability that must be overcome when using virtual computers. These must be addressed before ecommerce vendors will transition from physical servers to virtual ones. We had to solve these hurtles for our own geospatial exchange application (http://www.weogeo.com) developed on EC2, and have recently begun to transition this intelligent management software to the greater EC2 community (http://www.weoceo.com).
    paulbissett
  • dangerous idea

    Sorry, my enterprise's data presevation doesn't need the extra wildcard of Google's continued solvency, especially in light of thei recent purchase of youtube. Lawsuit city. I don't want a piece of that action, thanks.
    zandar
  • What about bandwidth costs?

    The concept is intriguing, but a real solution will also incur bandwidth and storage costs. It's also not yet clear what kind (speed) of connectivity there is between the multiple servers. i.e. will your web-server be able to access your DB server as if it were on a LAN?

    But this sure seems like the future of hosting.

    (Note to the previous poster, EC2 is currently an Amazon, not Google product, but you never know...)
    jmargolese
  • Re-Revolution you mean...

    This is simply an extension of an idea that pioneered by the mainframes of yesteryear, and a more recent example is Sun Solaris 8 on a SunFire 3800 (fully loaded), that can be clustered or operate in a single server mode. When fully outfitted with a full compliment of redundancy and capacity these machines, albeit incredibly expensive machine to purchase upfront, were made economical by multiuser/multiserver support. Solaris when coupled with server hardware allowed admins to remotely create multiple host domains dynamically as needed, and redistributed memory, hard drive space, and processing power as needed, and each "host" was independent. These multiple domains pooled resources in clouds on the system, such as 10 I/O network ports, 8 CPU's, and 64GB, a full 8 bay RAID array as dynamically configurable, shared memory/hard drive space. The commands were issued remotely, and outside the primary server hardware by a controller via a simple terminal connection. Thus dynamically created or collapsed hosts as well as how much capacity CPU/memory/HDD space would all be created distributed and removed whenever the system required. Now add the clustering interoperability, and extend it to support a front end for customers, and 'Poof', EC2.

    The use of an "API" command to the controller interface, represents the same method as EC2, and also exposes the consumer to a significant risk which wasn't addressed in the video, though I'm not sure that was the intent for the video anyways. At issue in terms of threat and tracking, eventually must be addressed in any system, but one with "virtual servers" I feel would be incredibly vulnerable. Drop a physical server, offline it, and you still can determine what happened with proper controls and monitoring via a forensic analysis. Drop a virtual server, what kind of forensic information does it provide, and is there any permanence to the information that was stored via these virtual machines on a third party cluster? Did the original data become deleted upon " instantiation" of the servers?

    So, the concept is not revolutionary, or very new at all, but another update of a previous concept and idea as a result of the shift to commodity based computing of the last decade. Simply returning to the concept of shared memory/processing systems and leveraging the incredible scale of Amazon's overcapacity in their server farms/co-locations, that they need to find another revenue stream that was simply considered an operating expense.

    Many questions, but the concept is good, and it's good buzz for Amazon to sell and get a return from their unused capacity and leverage their extensive infrastructure experience to capture revenue that otherwise would just have been lost.



    That being said, I can see the value for a small business, which the editor mentioned his own business use. However, when the "servers" fail, and his entire business is lost because of the presumed safety of the "value broker" of server space can't recover any of his files due to the systemic failure, I'm not quite sure if that savings will matter much to the future of his business.
    Mobious
  • are you really running enough traffic

    to justify paying Google $x/hour? Are you in the habit of getting slashdotted?

    A <$5/month low-end web hosting service might be a lot cheaper. I'd be looking at google "cloud" hosting if I were getting LOTS of traffic every day.
    A.Lizard
  • Call me simple minded but

    I don't know call me simple. But I think there is alot of value in the EC2 solution. The Internet and the items hosted on it are not static. Indeed most times they are very volitile. I think it would not only be fisically sound to seek a service that can be tweaked "On" and "Off" as demand calls for it. And you only be charged for that demand time. It seems a logical evolution or next step in web hosting.
    techweeklyblog