[ome-users] Hardware Requirements

[Ilya Goldberg]

On Jul 7, 2006, at 2:27 AM, InCho Chong wrote:

> -llya,
>
> Thanks for your response.  I am hoping to get some help from users who
> have tried installing OME on Red Hat Enterprise.
>
> On 7/6/06, Ilya Goldberg <igg at nih.gov> wrote:
>>
>> The data server has more complex requirements because it does a lot
>> of different things.  I/O is not as big an issue as CPU and RAM.
>
> Is this data server different than the OMEDS?

This is the OMEDS (DS == Data Server).

> Design document states
> that this is metadata storage, but I was not sure if this server will
> contain the image files also.

It can have them if you want - you can run OMEIS and OMEDS on the  
same machine, but they are also designed to run on separate  
machines.  The image files are always handled by OMEIS regardless of  
the setup.

> Is there a omeds and another data
> storage server using postgres?  Or is this the same server?

Postgres is yet another server - you can run the postgres DB on the  
same machine, but OMEDS can also connect to a remote postgres  
server.  The postgres server is just the standard postgreSQL -  
there's nothing OME-specific about it.  OMEDS talks to PostgreSQL.   
OMEIS doesn't talk to postgres at all - it just deals with image data  
by processing requests from OMEDS or other clients (like web browsers).

For example, when you are looking at a web page with thumbnails, the  
web-page is generated by OMEDS (technically, its web-ui component).   
The web page includes <img> tags for the thumbnails which request the  
images from the appropriate OMEIS.

>
>> If you provide more detail about the granularity of your data, how  
>> many
>> concurrent users you expect, etc, maybe I or someone else could chime
>> in with more specifics.
>
> All I am aware currently is that we hve about 2 terabytes of pixel
> data with a 100 MB dataset of TIFF files only differing in a couple of
> microscopic acquisition parameters but shares dozens of common
> biological attributes.  Not much more than that. Do we have to import
> all these files into the data server?

The data server is in charge of the import process, but the images  
and pixels end up on OMEIS.  The ome import command presumes that the  
image files are on the local file system.  They get sent to OMEIS via  
an http protocol wherever OMEIS happens to be - same machine,  
different box on the same rack, or half way around the world.  After  
they get sent to OMEIS, the Import Engine on OMEDS determines the  
format and extracts meta-data by talking to OMEIS, then issues  
"Convert" commands to OMEIS to extract pixels from the images.  OMEIS  
now has the original file, the extracted cached pixels, and some  
computed statistics.  OMEDS has the meta-data and a reference to the  
Pixels and the original file they came from on OMEIS (the reference  
is the OMEIS URL and a 64-bit integer ID).  Whenever various pieces  
of OME software (or independent libraries for C and MATLAB) need  
pixels, they get them from OMEIS in the same exact way regardless of  
the underlying image format.
-Ilya



>>
> --
> -InCho
>