JSAGA

Hi Patricia,
The purpose of a data adaptor is to manage (read/write/delete) files and directories on a storage area that can be either a grid storage element or a local file system.
Regards,
Lionel

Hi Patricia,

Indeed, the primary purpose of data adaptor is not related to job management.

However, it is also used by JSAGA to do the part of job staging that is not supported by the job adaptor. For example, if you want to retrieve WMS job output on your host, then 1st part of staging (from worker node to GSIFTP server) is done by WMS, while the 2nd part of staging (from GSIFTP server to your host) is done by JSAGA with the gsiftp data adaptor. This is transparent to the end user, but adaptor developers need to understand this.

Moreover, middleware rarely support both staging and I/O streams natively. JSAGA emulates the missing feature with the supported one (i.e. it emulates staging with I/O streams, or emulates I/O streams with staging). This is also transparent to the end user.

As an adaptor developer, the important thing to understand is that job adaptors should implement only the features that are natively supported by targeted middleware, and let JSAGA emulate the missing features.

Regards,
Sylvain

Hi all,

Right, but if I want to send the files required to run a job and retrieve the job output. Can I do that only by implementing the job control adaptor optional functionalities? or Do I need to implement the data adaptor?

Thanks in advance,
Patricia

Hi Patricia,

Your job adaptor should implement only the functionalities that are natively supported by the targeted middleware.

If your targeted middleware is able to transfer files between worker nodes and an intermediary storage server, but you need to transfer files between worker nodes and your local host, and there is no existing data adaptor for the protocol supported by this intermediary storage server, then you would need to implement this data adaptor.

Regards,
Sylvain