Re: [ecf-dev] Using ECF to create a osgi cloud...

[Cristiano Gavião <cvgaviao@xxxxxxxxx>]

Hi again,

answers below...

On 29/01/13 18:23, Scott Lewis wrote:
> Hi Chris,
>
> On 1/29/2013 5:22 AM, Chris Hawk wrote:
> > Hi Scott,
> >
> > thanks for the contextualization.
> >
> > It seems that I will have to research the possibilities by myself :)
>
> I wouldn't put it exactly this way.  Rather, I would say there are 
> several of us in the community that have some experiences...with 
> different providers with different use cases...that we can record and 
> share to our mutual benefit.
What I write was that I will do the investigation about the components 
used in providers myself since I haven't found any doc and don't receive 
any tip about them... but off course I will hope for experiences. ;)
> > I planed to create a pilot project that should be hosted in a 
> > OpenShit cloud free account. I will write a paper with what I found. 
> > I'll sent to you when I finish, so it could help you to format the 
> > wiki that you have created...
>
> You are certainly free to do this, but I would like to encourage you 
> to do this in the context of the ECF project.  We/I will agree to take 
> any/all code contributions and add them to our test suite (and/or 
> examples suite), and we can all add analyses and use case information 
> to the wiki pages.
>
> I do believe that several of us (myself, certainly...but also folks 
> like Pablo, Markus, Wim, Harshana and others) can/could contribute to 
> this over time, so my hope and belief that it would save you a fair 
> amount of work (as well as benefit the ECF community).

Yes, it is in context of ECF, but to achieve my goal the project that I 
will create needs to be deployed into a cloud environment.

> > btw, how do you suggest to measure performance and to track 
> > non-functional characteristics?
>
> WRT performance...with very few exceptions the performance of 
> discovery and distribution providers is bound by the pattern and costs 
> of network communication...e.g. how many connects, how many round 
> trips, how much overhead imposed by marshaling/unmarshaling. I think 
> the most effective measurements there are 'on-the-wire' costs (e.g. 
> via wireshark monitoring), along with timings of multiple remote 
> invocations...done multiple times obviously so that temporary 
> network-induced error can be mitigated.
>
> WRT non-functional characteristics (e.g. security, performance, and 
> reliability, etc).  I would suggest that we add sections for each 
> non-functional characteristic of important/interest, and then for that 
> section we add use case information (e.g. need remote service 
> discovery over wan, internet, etc).  And add info about the various 
> providers relative to that use case information.
>
> Again...although I/we will welcome any contributions, I hope you will 
> consider working closely with me and other ECF committers and 
> contributors...to minimize your work to collect this information, but 
> also to get community input (of which there is/will be more). If 
> appropriate and desired, we can also look into committership for you.  
> I will personally attempt to make this a priority over the next six 
> months, as I do think it will be very useful for the ECF community.
sure, I will be in touch as soon as possible.

> Thanks,
>
> Scott
>
>
>
>
> _______________________________________________
> ecf-dev mailing list
> ecf-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/ecf-dev