Sumatra and Docker
Felix Z. Hoffmann
for some time now I'm running computations from within a Docker container and track the simulations with Sumatra. I used Andrew's neuralensemble/simulation image as a base and built Sumatra on top.
If the correct ports are published (-p 8000) and smtweb is run with the option --allips, the web interface displays flawlessly on the docker host as well.
I haven't run into any real issues yet and believe that running the simulations from within a container is overall beneficial for replicability. For me it also solved a rather long standing issue regarding the mobility of Sumatra. Using the docker container, I'm able to run simulations and access records both from my work computer and personal laptop effortlessly.
The Sumatra images I created are on Docker Hub and I also wrote a full blog post on Sumatra and Docker. I would very much like to hear your thoughts on running and tracking simulations in a Docker container. Please share your experiences as well.
All the best,
Felix
Yannick Congo
Hello Felix,
Very nice work. I have been working on a platform since 2014. That will bring reproducible records to the cloud. CoRR (Cloud of Reproducibilr Records) is funded by MGI at NIST in Maryland (https://mgi.nist.gov/cloud-reproducible-records). One side of it is the integration of simulation management tools. The platform is open for regular Sumatra records up to container images, virtual machines images, etc.. (I would say solutions to preserve the execution environment). I would like to chat with you to compare our approaches. In my case of course CoRR take a over smtweb as a central cloud collaborative space. The design of the bare docker setup for scientist is at the heart of what I am building. How to make as easy to use as possible and sharable as possible. Which means having the data outside of the image so that people can try their own while recording the input and output files separately. Or bundling everything in the image and one can use external volumes. I picked the first one in my case for now. I think containers might be really suited for preserving the environment as maximum as possible. Yet there are situations where we cannot guaranty Reproducibility which is another of my goals. A set of articles on random numbers generators have proven that some generators using virtual machine images on a set of operating systems and machines were producing very different results. I would like to pick everyone's mind in that. I have been of course working into integrating CoRR api into Sumatra and planning later this year to address the ease and usefulness of having a cloud base collaborative space where Sumatra records can be used and allow the 'preservation' of the environment.
I am currently on a trip out of country for my engagement but couldn't go over your email :-) and not reply. I will be available by the end of February.
Thank you.
Yannick
Thank you
--
You received this message because you are subscribed to the Google Groups "sumatra-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sumatra-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.