River use cases

[Joerg Fritsch <fr...@joerg.cc>]

Hi,

what are currently the most appealing use cases for River? What are the best deminstrators that do show that River is what the world needs :D ?

--Joerg

Re: River use cases

[Joerg Fritsch <fr...@joerg.cc>]

Dan,

thsi is off topic of course, but can you pls send me the link to those videos? I attented some conferences etc. and I did not feel that amazon would have multicast enabled. I felt that the message was that at "exascale" they do not see multicast as feasible.

--Joerg


On Jan 5, 2013, at 12:18 PM, Dan Creswell wrote:

> Amazon != AWS :)
> 
> Internally, Amazon have multicast enabled, they went through a number
> of generations of it. There are some very old youtube presentations
> out there on this subject.
> 
> AWS is a completely different thing and doesn't have multicast.

Re: River use cases

[Dan Creswell <da...@gmail.com>]

On 5 January 2013 11:07, Joerg Fritsch <fr...@joerg.cc> wrote:
> Thanks for your response.
>
> I think my question should have been: "What else except web services can you do with it"? Eg distributed maps,caches etc. and how could / would I prrof the feasability or performance of those.
>
> I think that Gelertner may not have had exactly web services in mind when he first described LINDA & Tuple Spaces. So I look for applications that are _not_ web services together with easily accessible demonstrators. I once read that calculating atom orbital shapes or so and you would need an eight year study and atom physics to even code the demonstrator. I think I look for something more close to the "ordinary world".

Gelernter most definitely didn't have web services in mind. LINDA was
originally focused on shared-memory systems, in their most extreme
forms, hypercubes. He then described a version based on what he called
Trellises which was driving towards the asynchronous network (ethernet
and the like) world.

In Jini, a TupleSpace (JavaSpace in Jini parlance) is merely one
service available within the system. The TupleSpace model is generic
enough that it can handle maps or caching but the real power comes
from co-ordination in general and blackboard-style algorithms are
probably the most popular. Think workflow...

"how could / would I prrof the feasability or performance of those."

You'd write some appropriate benchmarks for that. I'm not sure but it
feels like you have a project of some form to do, perhaps for
university or work. If it's work, I'd expect you'd have some problem
in mind which I could help you formulate some tests for. If it's for
university, well, I'm not going to write your assignment for you ;)

>
> By the way: your comment conc Google and AWS I do not think that they have multicast enabled anywhere (for themselves), also multicast is not a feature that would be accessible to guest systems on the EC2 cloud.
>

Amazon != AWS :)

Internally, Amazon have multicast enabled, they went through a number
of generations of it. There are some very old youtube presentations
out there on this subject.

AWS is a completely different thing and doesn't have multicast.

Similarly for Google versus Google Engine.


> --Joerg
>
>
> On Jan 5, 2013, at 10:21 AM, Dan Creswell wrote:
>
>> River is basically an infrastructure for building service oriented systems.
>>
>> What can you build with a service oriented system? Pretty much anything.
>>
>> So what would the "most appealing use cases" be? Anything.
>>
>> Some things it does more naturally than others: Works well in Java,
>> less so with other languages. Of course that doesn't preclude doing a
>> non-Java front end and having it talk to Java and Jini services at the
>> back via RESTful interfaces and such. There are also some other means
>> (Surrogate) for working around this problem.
>>
>> It has a nice dynamic lookup mechanism so long as you have multicast
>> available in your networks. The likes of Amazon and Google do, for
>> some reason many smaller concerns don't.
>>
>> A standard 3-tier architecture with display, business logic and shared
>> database storage really doesn't fit well but then it wouldn't qualify
>> as an SOA so no surprise.
>>
>> If you have something more specific in mind, I can give you more
>> specific commentary but your question is so generic, so open, writing
>> anything with detail beyond the most superficial is impossible. Kinda
>> like "What's the performance of x?", lacking context.
>>
>> Hope that helps,
>>
>> Dan.
>>
>> On 4 January 2013 23:56, Joerg Fritsch <fr...@joerg.cc> wrote:
>>> Hi,
>>>
>>> what are currently the most appealing use cases for River? What are the best deminstrators that do show that River is what the world needs :D ?
>>>
>>> --Joerg
>>>
>

Re: River use cases

[Joerg Fritsch <fr...@joerg.cc>]

Thanks for your response.

I think my question should have been: "What else except web services can you do with it"? Eg distributed maps,caches etc. and how could / would I prrof the feasability or performance of those.

I think that Gelertner may not have had exactly web services in mind when he first described LINDA & Tuple Spaces. So I look for applications that are _not_ web services together with easily accessible demonstrators. I once read that calculating atom orbital shapes or so and you would need an eight year study and atom physics to even code the demonstrator. I think I look for something more close to the "ordinary world".

By the way: your comment conc Google and AWS I do not think that they have multicast enabled anywhere (for themselves), also multicast is not a feature that would be accessible to guest systems on the EC2 cloud.

--Joerg


On Jan 5, 2013, at 10:21 AM, Dan Creswell wrote:

> River is basically an infrastructure for building service oriented systems.
> 
> What can you build with a service oriented system? Pretty much anything.
> 
> So what would the "most appealing use cases" be? Anything.
> 
> Some things it does more naturally than others: Works well in Java,
> less so with other languages. Of course that doesn't preclude doing a
> non-Java front end and having it talk to Java and Jini services at the
> back via RESTful interfaces and such. There are also some other means
> (Surrogate) for working around this problem.
> 
> It has a nice dynamic lookup mechanism so long as you have multicast
> available in your networks. The likes of Amazon and Google do, for
> some reason many smaller concerns don't.
> 
> A standard 3-tier architecture with display, business logic and shared
> database storage really doesn't fit well but then it wouldn't qualify
> as an SOA so no surprise.
> 
> If you have something more specific in mind, I can give you more
> specific commentary but your question is so generic, so open, writing
> anything with detail beyond the most superficial is impossible. Kinda
> like "What's the performance of x?", lacking context.
> 
> Hope that helps,
> 
> Dan.
> 
> On 4 January 2013 23:56, Joerg Fritsch <fr...@joerg.cc> wrote:
>> Hi,
>> 
>> what are currently the most appealing use cases for River? What are the best deminstrators that do show that River is what the world needs :D ?
>> 
>> --Joerg
>> 

Re: River use cases

[Dan Creswell <da...@gmail.com>]

River is basically an infrastructure for building service oriented systems.

What can you build with a service oriented system? Pretty much anything.

So what would the "most appealing use cases" be? Anything.

Some things it does more naturally than others: Works well in Java,
less so with other languages. Of course that doesn't preclude doing a
non-Java front end and having it talk to Java and Jini services at the
back via RESTful interfaces and such. There are also some other means
(Surrogate) for working around this problem.

It has a nice dynamic lookup mechanism so long as you have multicast
available in your networks. The likes of Amazon and Google do, for
some reason many smaller concerns don't.

A standard 3-tier architecture with display, business logic and shared
database storage really doesn't fit well but then it wouldn't qualify
as an SOA so no surprise.

If you have something more specific in mind, I can give you more
specific commentary but your question is so generic, so open, writing
anything with detail beyond the most superficial is impossible. Kinda
like "What's the performance of x?", lacking context.

Hope that helps,

Dan.

On 4 January 2013 23:56, Joerg Fritsch <fr...@joerg.cc> wrote:
> Hi,
>
> what are currently the most appealing use cases for River? What are the best deminstrators that do show that River is what the world needs :D ?
>
> --Joerg
>

Re: River use cases

[Greg Trasuk <tr...@stratuscom.com>]

Essentially, it's an implementation technology for service-oriented
architectures, without all the usual XML-based complications that
traditionally go with SOA.

I wrote up more on this opinion here:
http://wiki.apache.org/river/JavaBasedSOA

Greg.

On Fri, 2013-01-04 at 18:56, Joerg Fritsch wrote:
> Hi,
> 
> what are currently the most appealing use cases for River? What are the best deminstrators that do show that River is what the world needs :D ?
> 
> --Joerg
>