|
|
Re: Concering bug 144950 [message #96182 is a reply to message #96125] |
Wed, 18 April 2007 03:18 |
Marius Slavescu Messages: 67 Registered: July 2009 |
Member |
|
|
Hi Holger,
In the log case we used our own EMF to RDB mapping framework (which was
build before CDO and Teneo).
That framework is generic and we already enabled it for Test (for execution
results, this function will probably surface in a post 4.4 tech preview)
model and I also did tests with the Trace and Statistical models (although
this will not be part of TPTP 4.4 GA).
The idea of service oriented data management infrastructure (TPTP DMS) is to
provide canned services to cover most TPTP UI/reporting/analysis uses cases
and also allow extensions and customizations for different data domains or
persistence mechanism.
An important feature of the new TPTP DMS is to provide a more flexible in
memory model that would be easy to page in/out and protect the client code
from memory problems (which we hit by using EMF).
Another idea is to use specialized data structures instead of just modeled
POJOs to optimize the data query and retrieval from the underlying data
store.
This DMS framework is inteded to be shared/reused within COSMOS project.
Bug 144950 and it's dependant bugs are intended to make the Test
UI/reporting components scalable by using queries instead of direct
navigation of the model and by using SWT virtual tree/table support combined
with other navigation helpers (we already did this for the Log UI and
partially in some Trace views).
Just having better scalability in the model is not enough we always need to
make sure that the UI/reporting/analysis components will leverage the model
scalability in the right way.
Regards,
Marius
"Paul Slauenwhite" <paules@ca.ibm.com> wrote in message
news:f02bie$1mi$1@build.eclipse.org...
> Hi Holger,
> Great news. We are starting to integrate the new service oriented data
> management infrastructure
> (https://bugs.eclipse.org/bugs/show_bug.cgi?id=169353) with the Test
> Project (https://bugs.eclipse.org/bugs/show_bug.cgi?id=177041) to attempt
> to resolve these scalability issues.
>
> Paul
> "Holger Machens" <machens@informatik.fh-wiesbaden.de> wrote in message
> news:ejkq71$qg1$1@utils.eclipse.org...
>> Hello,
>>
>>
>> I even saw bug 144950, were you are discussing the scalability of the log
>> framework.
>>
>> In my opinion, the scalability problem exists also for the trace model
>> and the test model.
>>
>> Arguments for the trace model might be obvious. For the test model,
>> consider a test suite
>> for replay of larger log files. According to the TPTP ASQ method, you
>> start with
>> an import of a record (i.e. the log) into a representation of the test
>> model:
>> This will need a similar amount of memory like the representation in the
>> log model.
>>
>> Hope, this is helpfully.
>>
>>
>> Greetings
>> Holger
>>
>>
>> BTW: Didn't you solve the problem for the log model by the usage of the
>> EMF database feature CDO?
>
>
|
|
|
Re: Concering bug 144950 [message #96514 is a reply to message #96182] |
Fri, 20 April 2007 14:15 |
Holger Machens Messages: 57 Registered: July 2009 |
Member |
|
|
Hi Paul and Marius,
thanks for your information and clarification!
Very interesting!
Regards,
Holger
Marius Slavescu schrieb:
> Hi Holger,
>
> In the log case we used our own EMF to RDB mapping framework (which was
> build before CDO and Teneo).
>
> That framework is generic and we already enabled it for Test (for execution
> results, this function will probably surface in a post 4.4 tech preview)
> model and I also did tests with the Trace and Statistical models (although
> this will not be part of TPTP 4.4 GA).
>
> The idea of service oriented data management infrastructure (TPTP DMS) is to
> provide canned services to cover most TPTP UI/reporting/analysis uses cases
> and also allow extensions and customizations for different data domains or
> persistence mechanism.
>
> An important feature of the new TPTP DMS is to provide a more flexible in
> memory model that would be easy to page in/out and protect the client code
> from memory problems (which we hit by using EMF).
>
> Another idea is to use specialized data structures instead of just modeled
> POJOs to optimize the data query and retrieval from the underlying data
> store.
>
> This DMS framework is inteded to be shared/reused within COSMOS project.
>
> Bug 144950 and it's dependant bugs are intended to make the Test
> UI/reporting components scalable by using queries instead of direct
> navigation of the model and by using SWT virtual tree/table support combined
> with other navigation helpers (we already did this for the Log UI and
> partially in some Trace views).
>
> Just having better scalability in the model is not enough we always need to
> make sure that the UI/reporting/analysis components will leverage the model
> scalability in the right way.
>
> Regards,
> Marius
>
> "Paul Slauenwhite" <paules@ca.ibm.com> wrote in message
> news:f02bie$1mi$1@build.eclipse.org...
>> Hi Holger,
>> Great news. We are starting to integrate the new service oriented data
>> management infrastructure
>> (https://bugs.eclipse.org/bugs/show_bug.cgi?id=169353) with the Test
>> Project (https://bugs.eclipse.org/bugs/show_bug.cgi?id=177041) to attempt
>> to resolve these scalability issues.
>>
>> Paul
>> "Holger Machens" <machens@informatik.fh-wiesbaden.de> wrote in message
>> news:ejkq71$qg1$1@utils.eclipse.org...
>>> Hello,
>>>
>>>
>>> I even saw bug 144950, were you are discussing the scalability of the log
>>> framework.
>>>
>>> In my opinion, the scalability problem exists also for the trace model
>>> and the test model.
>>>
>>> Arguments for the trace model might be obvious. For the test model,
>>> consider a test suite
>>> for replay of larger log files. According to the TPTP ASQ method, you
>>> start with
>>> an import of a record (i.e. the log) into a representation of the test
>>> model:
>>> This will need a similar amount of memory like the representation in the
>>> log model.
>>>
>>> Hope, this is helpfully.
>>>
>>>
>>> Greetings
>>> Holger
>>>
>>>
>>> BTW: Didn't you solve the problem for the log model by the usage of the
>>> EMF database feature CDO?
>>
>
>
|
|
|
Powered by
FUDForum. Page generated in 0.03098 seconds