You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@tuscany.apache.org by "Millies, Sebastian" <Se...@ids-scheer.com> on 2010/09/20 10:11:57 UTC

Q: relation of nodes to JVMs, interaction patterns

Hello there,

I have two elementary question about interaction patterns in SCA.
In section 4.1 of “Tuscany SCA in Action”, it says:

<quote>
Remote – The interacting components are running in two different Tuscany nodes, in
the same or separate Java virtual machines (JVMs), and could be running on different
physical computers. An SCA binding such as web services or JMS is required to allow
the components to communicate. Data passed between remote components is passed
by value.

Local – The interacting components are running in the same node and JVM. The
Tuscany runtime uses in-memory communication in this case. Data passed between
local components is passed by reference.
</quote>

First question with regard to remote interaction:
I work with JDK 1.6 on Windows XP. Any time I start a node through any of the
APIs (NodeLauncher, SCANodeFactory or SCADomain) I see a new Java process (JVM)
started up. Also, when I want to start a remote node in a distributed domain I
will use java.exe on the command-line. So there seems to be a one-to-one correspondence
between nodes and JVMs, contrary to the assumption in the above quote. Or is there an 
API that will simply start a new thread and use a separate classloader to create a new 
node, or something? Or is this operating system dependent? If so, in what system, and how, 
could I set up two separate nodes running in the same JVM and interacting remotely?

Follow-up question with regard to local interactions: If two components communicate 
through local interfaces only, is there anything from a design point of view against 
passing normal Java references as method parameters, e. g. "addListener(this)" as opposed 
to "addListener(componentContext.createSelfReference(myLocalInterface.class))" ?

-- Sebastian

Re: Q: relation of nodes to JVMs, interaction patterns

Posted by Simon Laws <si...@googlemail.com>.
Hi Sebastien

Some comments in line...

Simon

On Mon, Sep 20, 2010 at 9:11 AM, Millies, Sebastian
<Se...@ids-scheer.com> wrote:
> Hello there,
>
> I have two elementary question about interaction patterns in SCA.
> In section 4.1 of “Tuscany SCA in Action”, it says:
>
> <quote>
> Remote – The interacting components are running in two different Tuscany nodes, in
> the same or separate Java virtual machines (JVMs), and could be running on different
> physical computers. An SCA binding such as web services or JMS is required to allow
> the components to communicate. Data passed between remote components is passed
> by value.
>
> Local – The interacting components are running in the same node and JVM. The
> Tuscany runtime uses in-memory communication in this case. Data passed between
> local components is passed by reference.
> </quote>
>
> First question with regard to remote interaction:
> I work with JDK 1.6 on Windows XP. Any time I start a node through any of the
> APIs (NodeLauncher, SCANodeFactory or SCADomain) I see a new Java process (JVM)
> started up.

I would probably expect new JVMs to be started when nodes are started
from the domain manager
but not in other circumstances. For example, see
samples/calculated-distributed. If you run this sample with Maven
it runs several nodes in the same JVM (if you run is with Ant you run
the nodes in separate JVMs). I do see two JVMs in the maven case but
one is for running Maven and the second is running Surefire which is
turn runs the sample.

>  Also, when I want to start a remote node in a distributed domain I
> will use java.exe on the command-line. So there seems to be a one-to-one correspondence
> between nodes and JVMs, contrary to the assumption in the above quote. Or is there an
> API that will simply start a new thread and use a separate classloader to create a new
> node, or something? Or is this operating system dependent? If so, in what system, and how,
> could I set up two separate nodes running in the same JVM and interacting remotely?

You should just be able to create a node factory and from that create
nodes as required...

            SCANodeFactory nodeFactory = SCANodeFactory.newInstance();
            nodeC  =
nodeFactory.createSCANodeFromURL("http://localhost:9990/node-config/NodeC");

This is from the distributed calculator so it's pulling the node
configuration (which contributions to load and which composite to run)
from a URL in this case. There are several createNode type methods
which allow you to provide this configuration in different ways.

>
> Follow-up question with regard to local interactions: If two components communicate
> through local interfaces only, is there anything from a design point of view against
> passing normal Java references as method parameters, e. g. "addListener(this)" as opposed
> to "addListener(componentContext.createSelfReference(myLocalInterface.class))" ?

Local interaction will follow a pass by reference approach if the
service interface is not marked as remotable. If the service interface
is remotable, i.e. it's a Java interface and has th @Remotable
annotation or it's a WSDL interface, then
parameters will always be passed by value regarless of whether the
reference and service are in the same JVM or not

Having said this, whether you pass an SCA service reference or a
normal Java reference into a service operation will depend on the
parameter type of the operation. createSelfReference() will return a
ServiceReference<B>. You could take the ServiceReference and call
getService() on it to get a proxy of the Java type. The question then
becomes should you pass the proxy or "this"? If you pass the proxy
then the SCA runtime handles and calls to the service and gets to
apply any policy that's been configured for the service. If you pass
"this" then that's not the case.

>
> -- Sebastian
>