[AG-TECH] AccessGrid 3: What information is available

[Ivan R.Judson]

I believe (from what I've seen) the move away from GSITCP increases  
performance *HUGELY*; it's an entirely different system  
performance-wise.

configuration mgmt is something I know has been talked about, but I'll  
let Tom or someone else from the Team answer.

--Ivan

On May 24, 2005, at 9:38 AM, Derek Piper wrote:

>
> 	While we're on the subject of globus and stuff like that, will we  
> still have to wait days in order to get a node up and running because  
> of getting the certificates? It's one of the most tiresome things  
> about setting up a new node.
> 	Also, something I'd REALLY like to see is the ability to set a  
> site-wide configuration under Linux and Windows. Sure have the users  
> have a .AccessGrid folder (or the equivalent under 'Documents and  
> Settings' for Windows) but having a system-wide config would be VERY  
> advantagous to me in setting up conference room AG nodes. Then, a user  
> known to the network can log in and it's already set up with the same  
> config everyone else uses. I don't like having to have just one user  
> run the AG stuff, especially with shared apps and the possibility that  
> new users want to get on the AG for a meeting, share their stuff in a  
> conference and get off. It would be nice if they could use a site  
> conference room and fairly much do it themselves, albeit with a  
> modicum of training.
>
> 	Derek
>
> John Hodrien wrote:
>> On Tue, 24 May 2005, Ivan R.Judson wrote:
>>> Hey John,
>>>
>>> Good point. Although my skepticism keeps nagging with the worry,  
>>> "Why bother
>>> with WSRF/GT4 services, aren't Web Services good enough?"  I suspect  
>>> they
>>> are, I haven't seen any significant value to the layers above that  
>>> provided
>>> by GT4.
>> If you try and ignore the globus stuff, and think of it as WSRF it  
>> becomes
>> more pleasant.  GT4 contains lots of crap for submitting jobs and  
>> managing
>> resources, copying files about etc.  I have no immediate interest in  
>> any of
>> that.  WSRF is the OASIS working draft for stateful web services, and  
>> as such
>> is intended to be something nice and generic, usable whenever you  
>> want state
>> in a web service.
>> The most important part of it as I'm concerned (if you were  
>> entertaining the
>> notion of AG with web services) would be WS-Resource and  
>> WS-Notifications.
>> http://docs.oasis-open.org/wsrf/2005/03/wsrf-WS-Resource-1.2-draft 
>> -03.pdf
>> http://docs.oasis-open.org/wsn/2004/06/wsn-WS-BaseNotification-1.2- 
>> draft-03.pdf Sadly you can quickly complicate things, as you're  
>> likely to need
>> WS-Reliability if you're going down the notifications road.
>> http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=wsrm
>> I'm not sure what else would be worth poking at, bits like  
>> WS-ResourceLifetime
>> might also deserve a look in.
>> Equally, playing with WS-Security and using the SAML callout would  
>> let you
>> define whatever you wanted on the security front, and make it open and
>> standardised how you were doing it.
>> These are all boxes of tricks that live independently of GT4  
>> (WSRF.NET for
>> example) but would open up the AG to open and standard interaction  
>> with other
>> software.
>> So really my case for this isn't in favour of GT4, it's in favour of
>> standardised (very nearly at least ;) stateful web services.
>>> Did I miss something :-)?
>> I'm sure we've all missed plenty ;)
>> jh
>
> -- 
> Derek Piper - dcpiper at indiana.edu - (812) 856 0111
> IRI 323, School of Informatics
> Indiana University, Bloomington, Indiana
>