[ome-devel] Who's who in Pixels? (iRODS ideas...)

[Harri Jäälinoja]

Hi Josh,

>> Hmm, maybe in this scenario it would anyway be useful to add an iRODS metadata field to the data set, to tell which OMERO server has the metadata for it. So when you register a data set from a collaborator at your OMERO site, the steps would be:
>
> That certainly seems plausible. Or, if iRODS supported "federated metadata" it could just proxy all the metadata from OMERO (I have *no* clue if this is possible or not)
>

the iRODS metadata consists of (attribute, value, unit) triplets (AVUs) 
that are stored in the database. If these were to be used, it would 
first be necessary to get the metadata from OMERO and populate the iRODS 
database. I don't think it would be possible to get the metadata 
directly from OMERO in a transparent/federated way.

There is also a system to access external databases:
https://www.irods.org/index.php/Database_Resources
It sounds like with this you could access the OMERO Postgres database 
and query metadata.

Federation in iRODS means integration between zones 
(https://www.irods.org/index.php/Federation). A zone always has it's own 
metadata catalog. It seems that in the standard cross-zone replication, 
metadata would not be replicated (http://code.google.com/p/extrods/), 
and with the CZDRS code you can replicate also the metadata.

Here's also a presentation where on pages 33-39 some federation 
deployment models are outlined:
http://colab.mpdl.mpg.de/mediawiki/File:ESci_09_Sem_2_iRODS_Federation_Hasan.pdf

>> # register and tell where to look for metadata
>> bin/omero register --omerohost=$OMEROHOST $DATASET
>>
>> Or something like that :)
>
> :) Seems like that last command would still end up creating two separate copies in OMERO. I guess it all depends on what one wants to do with the second iRODS replicate. If it's to launch, say, OMERO scripts, then there will need to be some way to go about getting the script to one of the replicates. Hmmm....
>

Here my fantasy was that you have local OMERO instances at collaborating 
institutes A and B. Both A and B also use the same iRODS data grid. When 
people at A want to analyze data from B, they can use iRODS to make a 
local replicate of the data on a drive where it will be visible to 
OMERO. Then they would register it, and they would be able to run 
scripts. And my fantasy continued so that now OMERO at A knows that it 
is not the rightful owner of the data set, and when users make 
annotations etc., it will forward them to OMERO at B. OMERO at A would 
start to act as proxy to OMERO at B, for this data set. I've no clue if 
there is even the tiniest possibility this might not remain a fantasy :)

If the institutes A and B have their images on a federated iRODS data 
grid, then they can see all of the data as one big virtual collection. 
And if they have access rights, they can read the data. I guess the 
people who have been asking the OME team about possibility to manage 
data between institutes might be looking for the same in OMERO; a way to 
browse projects and datasets in OMERO, even if the data is at different 
locations. Since I'm already getting used to iRODS, I would be content 
to use it for locating the data and transferring it wherever I need. For 
example, the project and dataset IDs can be stored in iRODS metadata, so 
then it will not be too difficult to find the correct data set. And when 
I need to actually see the images, run batch analysis and so on, I would 
use OMERO.

It's a nice puzzle :)

Cheers,
Harri