Large(r) scale installation

[mgruenb]

Hi everyone,

apologies for yet another installation question. I have read all the docs and searched the forum so I roughly know what we should be doing but it would be great to get some feedback from the OMERO community on our plans.

Quick background info: At Cambridge Uni we've been using Hamamatsu NDP for a couple of years for teaching around 250-300 undergrads, the server software isn't great so we're now planning on using OMERO. I'm doing this as a side-project so once it's all set-up I'd like to keep server admin to a minimum. Images are the usual, mostly 100s of MB to a 4-5 GB for some.

The plan + questions:

Server 1 with big TB HDs: File server, let users drop images via Windows shares (SMB) and use the OMERO dropbox feature. Make images available to OMERO servers 2+3. Can you recommend the best way? NFS, SMB, other?

Servers 2+3: OMERO servers incl Postgres on each. One will be OMERO master the other OMERO slave. Also use Postgres in a master - slave set-up. Would that work? Or would it be easier to just rsync server 1 to server 2 regularly? If this set-up is too slow for teaching we might want to add more servers to this layer or perhaps rsync the most frequently used images to the local FS.

Server 4: Load-balancer/ cache/ proxy thingy. Maybe HAProxy? Any recommendations welcome

Ideally the images are then just available to anyone and we can link to them from the rest of the teaching material (a Wordpress system) and I don't have to administer the OMERO system at all for now

Not essential, but what's the best way to get annotations onto the images? Can we use the old NDPI XML files? Or should we do it in some other OMERO way? There aren't loads at the moment so we might as well start with the best way now.

Thanks,

Michael.

[jmoore]

Hi everyone,

Hi Michael,

apologies for yet another installation question. I have read all the docs and searched the forum so I roughly know what we should be doing but it would be great to get some feedback from the OMERO community on our plans.

No worries, that's what we're here for...

Quick background info: At Cambridge Uni we've been using Hamamatsu NDP for a couple of years for teaching around 250-300 undergrads, the server software isn't great so we're now planning on using OMERO. I'm doing this as a side-project so once it's all set-up I'd like to keep server admin to a minimum. Images are the usual, mostly 100s of MB to a 4-5 GB for some.

... and in fact, more and more people are looking to set up something like this, so your experience(s) will be invaluable.

The plan + questions:

Server 1 with big TB HDs: File server, let users drop images via Windows shares (SMB) and use the OMERO dropbox feature. Make images available to OMERO servers 2+3. Can you recommend the best way? NFS, SMB, other?

If possible, run a SMB server with access to the physical file server. OMERO.DropBox requires OS-level file change notifications to work properly. Then, these files can either be imported with "ln" (hard-linked) or with "ln_rm" (move). See http://www.openmicroscopy.org/site/support/omero5.1/sysadmins/import-scenarios.html for more information. In either case, you'll likely want to use NFS to export the filesystem to your other servers, but be careful of locking issues.

Servers 2+3: OMERO servers incl Postgres on each. One will be OMERO master the other OMERO slave. Also use Postgres in a master - slave set-up. Would that work? Or would it be easier to just rsync server 1 to server 2 regularly? If this set-up is too slow for teaching we might want to add more servers to this layer or perhaps rsync the most frequently used images to the local FS.

Keeping a Postgres slave up-to-date is certainly a very safe mode to work in. You'll still want to keep period backups of the FS with the DB if at all possible. Additionally, we have plans for adding a read-only mode to OMERO so that you could optionally host some content from the slave itself.

Server 4: Load-balancer/ cache/ proxy thingy. Maybe HAProxy? Any recommendations welcome

If most of your content is going to be read-only (for most users), then starting with heavy caching at the web server level (nginx/apache) is likely sufficient initially.

Ideally the images are then just available to anyone and we can link to them from the rest of the teaching material (a Wordpress system) and I don't have to administer the OMERO system at all for now

http://help.openmicroscopy.org/publish.html#repository has some tips & examples of setting up your system for publication. You'll certainly want to use the public user functionality for embedding URLs.

Not essential, but what's the best way to get annotations onto the images? Can we use the old NDPI XML files? Or should we do it in some other OMERO way? There aren't loads at the moment so we might as well start with the best way now.

It really depends on what types of annotations you're talking about. If the images are already imported, then likely a script which parses out further annotations and includes them in the OMERO DB will be optimal. If the images aren't imported yet, it might be worth our looking at one of the files with an example of an annotation that you'd like to see.

Thanks,
Michael.

Cheers,
~Josh.

[mgruenb]

Thanks for the reply Josh!

It all makes sense (and it'll keep me busy for a while)

I'll try and document everything properly and then I'll fill in your questionnaire and I'd be happy to share all details. I guess we'll be trying different options once students are back and I'll try and collect some benchmarks.

Cheers,

Michael.