Alan Williamson on Cloud Computing

Discussions

News: Alan Williamson on Cloud Computing

  1. Alan Williamson on Cloud Computing (7 messages)

    Alan Williamson takes on the hype and the reality of cloud computing in a java.sun.com interview(http://java.sun.com/developer/technicalArticles/Interviews/community/williamson_pt1_qa.html). Williamson is apparently the first person to run a cloud computing boot camp and now edits the new cloud computing journal. -- Williamson disputes the idea, touted by Richard Stallman and others, that cloud computing will be a threat to open source because it will lock people into proprietary software and that moving from the desktop and datacenter to the network cloud will shut out open source software. According to Williamson, the opposite will be true as people will demand checks and assurances that they can trust their infrastructure to the cloud provider. Does this make sense? Why do such assurances need to involve open source? -- Williamson makes monitoring and managing the cloud sound like a breeze thanks to JMX clients, like JConsole. But will it really be a breeze given the demands we can expect from cloud infrastructures? -- Williamson states: "...it's the little things that can really throw your project a curveball. For example, can your application cope with MySQL/JMS/caching servers suddenly dropping out and reappearing on a completely new IP address? These are the things you need to build." Does this ring true?
  2. Williamson states: "...it's the little things that can really throw your project a curveball. For example, can your application cope with MySQL/JMS/caching servers suddenly dropping out and reappearing on a completely new IP address? These are the things you need to build." Does this ring true?
    I am getting slightly angry reading a sentence like this. If *I* have to cope with such a situation, I do not need "the cloud" in the first place. And I definitively do not want to have to build something for coping with such a situation. I want the provider of the cloud infrastructure (whatever that is) to provide a defined and rock solid way that makes things like this transparent to the application developer! Next thing you're telling me, I would need to be aware what bus my hard disk is on and to prepare to cope with sudden disk failures of my SAN storage!
  3. Karl, what is being described here is simply hosting - not cloud computing. This as described is not going to lead to the cost savings people are expecting other than merely short-circuiting the server procurement/provisioning process. William
  4. I am getting slightly angry reading a sentence like this. If *I* have to cope with such a situation, I do not need "the cloud" in the first place. And I definitively do not want to have to build something for coping with such a situation. I want the provider of the cloud infrastructure (whatever that is) to provide a defined and rock solid way that makes things like this transparent to the application developer!
    I absolutely agree. There is nothing that should prevent you from running a standard JEE application on this type of environment without changing your code. The right approach should be to bring the value of cloud computing to the application rather then trying to bring the application to the cloud which seem to be some of the current way of thinking with some of the existing cloud framework. You can see here a live example showing how you can run a Pet Clinic application with load-balancer, web-container and MySQL as the backend database, this application is designed to scale-out dynamically when the workload goes up, all the issues of dynamic IP allocation, self healing is dealt with within the middleware layer. Nati S. www.gigaspaces.com/cloud

  5. I am getting slightly angry reading a sentence like this. If *I* have to cope with such a situation, I do not need "the cloud" in the first place. And I definitively do not want to have to build something for coping with such a situation. I want the provider of the cloud infrastructure (whatever that is) to provide a defined and rock solid way that makes things like this transparent to the application developer!


    I absolutely agree.
    There is nothing that should prevent you from running a standard JEE application on this type of environment without changing your code.

    The right approach should be to bring the value of cloud computing to the application rather then trying to bring the application to the cloud which seem to be some of the current way of thinking with some of the existing cloud framework.

    You can see here a live example showing how you can run a Pet Clinic application with load-balancer, web-container and MySQL as the backend database, this application is designed to scale-out dynamically when the workload goes up, all the issues of dynamic IP allocation, self healing is dealt with within the middleware layer.



    Nati S.
    www.gigaspaces.com/cloud
    Seemingly what you describe here amounts to: (1) Leave the application the same (2) Do all the hard work in the middleware layer (3) Leave the cloud the same Thus the cloud vendor is still requiring that the overall system be modified to fit with their infrastructure you're just attempting to contain it in the middleware as opposed to the application. Cloud computing platforms can only deliver their cost savings (via economy of scale) to users by adopting models that are not the norm in most enterprises. These models are all about no static assumptions, failure happening, boxes dying, high system to administrator counts (thousands to one) etc. These cloud platforms are typically an outgrowth of internal infrastructure developed with these new models because it's the only way Google, Amazon and co can achieve sustainability. They've built their infrastructure and applications in such a fashion as to keep cost (human intervention in provisioning, repairs and deployment etc) down. When they offer these platforms externally, should they have to do all the things expected by most enterprises their costs go up and of course they charge more eroding their market because they're closer to conventional hosting solutions. In summary, cloud offerings are different and deliberately so because it's how the cost savings are achieved. Customers wishing to take advantage of those cost savings either have to architect their applications differently or rely on someone else to provide an abstraction that paves over the cracks.
  6. In summary, cloud offerings are different and deliberately so because it's how the cost savings are achieved. Customers wishing to take advantage of those cost savings either have to architect their applications differently or rely on someone else to provide an abstraction that paves over the cracks
    Dan i agree with your assessment. The challenge that i see right now is that the gap between the existing cloud framework and the way Enterprise application are designed is too big and would become huge obstacle for greater adoption of cloud computing by those companies. My argument is that it doesn't have to that hard to bring enterprise application to the cloud. See the following example of one of the leading Telco providers who just recently launched a new Telco Enterprise Application on the cloud. Nati S gigaspaces.com/cloud
  7. What happened to this thread? There was 9 postings yesterday and not there is only 3.
  8. What happened to this thread? There was 9 postings yesterday and now there is only 3.