perfsonar development work

[Loukik Kudarimoti <>]

Nina Jeliazkova wrote:
> > Nina, has this helped you?
> Not really.
>
> The examples I sent this morning are as follows:
>
> 1)SetupDataRequest_FirstMeta.xml  - this has data linked to the FIRST metadata
> and  is working, i.e. it retrieves some data for the interface.  
>
> 2)SetupDataRequest_LastMeta.xml  - has data linked to the LAST metadata and is
> NOT working, i.e. the response does not contain data.
>
> The problem is that PerfsonarUI was communicating with the previous releases
> with requests similar to SetupDataRequest_LastMeta.xml  , the data linked to
> the LAST metadata of the chain.  However, this request does not work with the
> recent SQL MA. 
>
> There is no problem to change the philosophy from the client side and generate
> requests with data linked to the FIRST data for the new releases, but we have
> to be sure this is the right way. It is slightly confusing that the Interface
> specification says the data should be linked to the LAST metadata, which
> doesn't work with current RCs.  
> The Example No.2 from Interface specification of SQL MA also links data to the
> LAST metadata.
>   
Nina,

I think there is a mis-understanding here. I just spent some time 
testing out the SQL MA by making use of the documentation provided for 
the RRD MA.
In short, your requests are working as expected. Just try the second 
request (SetupDataRequest_LastMeta.xml) with the start time: 1143034700 
and the end time 1143035770

SQL MA RC6 is available here: 
http://mu.dante.org.uk:8092/perfSONAR-SQL-MA-1.0-RC6/services/MeasurementArchiveService

RC5 is no more!

Here are my findings:

In your first request (SetupDataRequest_FirstMeta.xml), the first 
metadata contains metadata paramters such as interface etc, which is 
used to find which data to retrieve. The second metadata contains filter 
paramters. By linking the data element to the first metadata element, 
you are not providing any select parameters (i.e. start and end time) 
because the second metadata is not in the chain. The service hence tries 
to retrieve data using default start and end times. The important point 
is that the default start and end times for SQL MA (utilisation data) is 
the last know utilisation value.

Hence, as a response for this request, you get the last known value.

In your second request(SetupDataRequest_LastMeta.xml), the select 
metadata is part of the chain so the service tries to find data for the 
given time interval. It so happens that for test purposes, during 
installation, there is just one data value entered into the sql database 
for one metadata. The time value for this data entry falls out of the 
range of the request that you are sending. Hence, as a response for this 
request, you do not get any values because there are no values in that 
time range.


Let me know if this sorts out the problem.

> Regards,
> Nina
>
>
>
>   
> > Roman just released RC6 for SQL MA. It has enhanced documentation. I 
> > will be installing it soon on our workstation. I will let you know when 
> > its done.
> > You can access its documentation here: 
> >
> >     
> http://anonsvn.internet2.edu/svn/perfsonar/tags/JAVA-SQL-MA-1.0-RC6/doc/ma/sqlTypeMA/
>   
> > Also, RRD MA RC8 installation is accessible here: 
> >
> >     
> http://mu.dante.org.uk:8092/perfSONAR-RRD-MA-2.0-RC8/services/MeasurementArchiveService
>   
> > Please let us know if you need more help. Its crucial for us to make 
> > sure that the visualization tools can work with the software that we 
> > release. Also, if you are not aware of it, we will from now on have 
> > version number - major and minor for perfSONAR bundles.
> >
> > Bundle Major version numbers - backwards incompatible protocol changes 
> > compared to previous versions
> > Bundle Minor version numbers - usually changes not related to protocol 
> > but if there are, they are usually backwards compatible.
> >
> > We hope that you can use these as a basis for your developments. If you 
> > have any questions let us know.
> >
> > Each service will have micro-release version numbers as well. But we 
> > expect these to be significant for the respective development teams but 
> > not for the developers of analysis tools.
> >
> > Andreas, what about CNM?
> >
> > regards,
> > Loukik.
> >
> >     
>
>
>
>