Use OMERO.server and Huygens Core on the same physical machine

Hi everyone,

At my workplace, we have some problems to acquire the necessary hardware to install our definitive OMERO server, so we examined the possibility to run OMERO.server on another workstation, potent enough to be used as a server. On this machine, Huygens Core (a server software for image deconvolution) is already installed.

My question is: Is it possible to run OMERO.server (with OMERO.web as front-end interface) and Huygens Core (with Huygens Remote Manager as front-end interface) on the same machine, in other terms the same IP address?

Best regards, Marc.

Hi @mmongy,

In a very vague way, yes, it’s possible to run OMERO with other processes on a machine. Things you might want to think about:

  • Several ports will either need to be free or be re-configured (OMERO’s 4064, PostgreSQL’s 5432).
  • If the server is already running a website on the standard ports (80/443), then you will need to “embed” OMERO’s web paths in your existing configuration.
  • You should separate each process by user.

~Josh

cc: @ehrenfeu in case he’s ever tried this

Hi @mmongy and @joshmoore,

I have never tried myself, but knowing both applications quite well I can comment on the following:

  • Huygens Core itself doesn’t open any ports nor have a database, so no collision expected here
  • in case you’re running the HRM web interface for Huygens Core, this is using a database (usually MySQL, but PostgreSQL is supported as well), but doesn’t put heavy load on the DB - and it’s (surprisingly) using a web server, but that can be the same that is also serving OMERO.web
  • there is the HRM-OMERO connector (you may guess who is the main contributor) that allows for some integration between OMERO and the HRM although I clearly have to say that it is not at all optimized for the case that both tools are sitting on the same machine (meaning Huygens would be able to access data in OMERO’s managed repository directly on-disk), but it’s still a convenient way for users to run deconvolution on data that is in OMERO and re-import the results
  • I second Josh’s suggestion to use different system users for OMERO and Huygens
  • you might have to test if renice’ing the Huygens process works in a reasonable way to make sure there are enough resources for OMERO, Postgres et. al. as otherwise system load might kill you (on our “old” HRM server with 24 cores we’re regularly seeing the load between ~20 and ~40)

Hope that helps!
~Niko

cc: @SVI_Huygens to double-check whether my statements about HuCore are correct.

1 Like

Hi everyone,

Yes, that’s indeed possible. There’s an HRM demo server hrm.svi.nl with exactly this setup. As a demo server it’s probably less heavily loaded than the average production, high-throughput machine but as far as we have seen it works quite nicely.
Plus, if the machine has GPU(s) then most of the image processing done by Huygens Core will be offloaded to it, so that should help too.

As for Niko’s comments on Huygens Core, yes, that’s all correct :slightly_smiling_face:

Here’s a link to the few steps you need to configure in HRM in order to enable communication with OMERO: https://huygens-remote-manager.readthedocs.io/en/latest/admin/configure_omero.html

I hope this helps.

Cheers,
Daniel Sevilla
For the SVI-Huygens Team

1 Like