JP Morgenthal wrote:
> The problem with moving the complexity into the infrastructure is that we
> have no trained resources capable of managing this complex infrastructure
> and understanding what the heck the developers are deploying into it.

Spec'd APIs such as Jini can, though, provide enough containment in 
features to focus the efforts of the developers into tractable designs. 
  The complete set of features that Jini provides cover the complete set 
of things that distributed, scalable applications need.  Certain 
applications need to make use of such features (Jini provided or 
otherwise) in specific ways to truely enable complete scaleability. 
There is no silver bullet distributed system implementation yet, which 
provides true autoscaling and QOS.  But, the RIO project, with its QOS 
features including auto redeployment(http://rio.jini.org) provides a 
fair amount of satisfaction to its users it appears.

> There's very little in the way of manageable and portable domains, such that
> demand for one application can pull from a resource pool and then return the
> allocation to the pool when complete.

The primary feature required of a distributed system which allows 
dynamic assignment of resources, is some means for recovery and tracking 
of those resources.  Jini's distributed leasing service and APIs along 
with the closely associated distributed transaction service make it 
possible to easily manage the resources of a distributed system.  Lost 
users of a resource will fail to renew leases and thus allow the 
recovery of those resources.  The 2P transaction manager allows mutual 
agreement between services to proceed to a concluded state for roaming 
resoures.

 > If you're gonna tell me that J2EE
> does this, you're off your rocker.  J2EE does 1/10th of what's really needed
> here to allow developers to develop simple application objects and deploy
> them into a generic field that is completely managed.  In fact, so many J2EE
> people I speak with agree that the promise of this has been unfulfilled by
> J2EE and that many early applications that bought into this are now
> undergoing re-writes to support the need for better scaling.  That is, the
> application had to be rewritten to scale, isn't this what the infrastructure
> was supposed to do automatically?

The J2EE design has many issues.  I appreciate your thought provoking 
comments here.

> Until we have an infrastructure that truly manages the complexity required
> by large-scale application and the trained resources to manage such an
> environment, the complexity must be shared across the application and the
> infrastructure.  This balance is what will make SOA difficult to really
> deliver over the next 10 years because the services will have to eventually
> absorb what the infrastructure cannot.

POJO is empowering.  The J2SE platform is very empowering when you can 
exploit all of its features.  Those that choose to ignore Jini are only 
destined to recreate its features at some point.  So I am still 
wondering why people don't choose to start out ahead and use the power 
of the Jini platform for their SOAs that use Java (or which could use 
Java and be even better).

Gregg Wonderly




 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/service-orientated-architecture/

<*> To unsubscribe from this group, send an email to:
    [EMAIL PROTECTED]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 


Reply via email to