We have three servers that can be used to run docker containers: puffer.cs.ait.ac.th, gourami.cs.ait.ac.th and guppy.cs.ait.ac.th. See the hardware configuration at the bottom of this page.

Using JupyterHub

Any user of CSIM and DSAI can access any of these servers through a JupyterHub interface to run Python notebooks. (At the moment, guppy is not available).

There is no further restriction.

Running containers manually

See important points about the migration of your existing containers and data.

Students enrolled in Machine Learning courses can also access the server directly via SSH and manually run their own containers. But security concerns arise when students run container under the default root username.

You must run your containers under your own username and groupname: To launch a docker container, user the option -u or --user in your command:

docker run -u $(id -u):$(id -g) ...

Container that run under root username are stopped automatically. 

As a direct consequence, your container will not be able to use docker volume. Instead you can use storage on your main home directory or in a dedicated directory in /home2.

Using your home directory

After you SSH'ed to one of the servers, use the command:

docker run -u $(id -u):$(id -g) -v `pwd`:/app ...

and your container will be able to save data in the directory /app, The data will be accessible from your home directory from any machine on CSIM network.

Using your home directory is slow because the data are stored on the network, but the data are readily accessible from any machine.

Data saved in your homedirectory will be limited by the quota on your account. Use the command quota -s or go to the account management page.

See important points about the migration of your existing containers and data.

Using a local directory in /home2

You can create a personal directory in /home2. It will be faster to use because it does not cross the network, but will only be accessible from that single server.

First step is to create the directory:

mkdir /home2/st123456

You must use your own username for the directory. Any directory with a name that does not correspond to a username will be made unavailable.

Then launch your container:

docker run -u $(id -u):$(id -g) -v /home2/st123456:/app ...

Similarly, your container will be able to save data in the directory /app and when not using a container, the data will be accessible in /home2/st123456.

There is no limit on the size of the data you can use, but the space is still limited to the size of the disks on the servers. Also consider that larger data may exceed the capacity of our backup.

See important points about the migration of your existing containers and data.

Migration

Even though the new policy will not be enabled immediately, you can start applying it from now-on. That may give you some amount of time to iron-out any problem.

There are few considerations you should be aware of when the new policy is enabled.

Migration of your containers

When the policy is enabled, containers currently running under root username will be allowed to continue until their termination. But you will not be able to run them again once they have stopped. 

Migration of your data

Data created before enabling the policy belong to the user root. You may not be able to access the data anymore. Please This email address is being protected from spambots. You need JavaScript enabled to view it. to correct the problem.

Container and data end of life

When your account with CSIM or DSAI expires, any of your container still running will be terminated automatically.

If the size of your data on the servers is not excessive, the data will be preserved, else, data created on the servers, via JupyterHub or in /home2 will be lost.

If the space in /home2 is being exhausted and other users cannot save data anymore, we may have to remove the data from the largest user.

Hardware of the servers

puffer	32 x CPU, 64GB memory, 4 x NVIDIA GeForce RTX 2080 Ti
gourami	8 x CPU, 32 GB memory, 2 x NVIDIA GeForce GTX 1080 Ti
guppy	8 x CPU, 15GB memory, 2 x NVIDIA GeForce GTX 1080 Ti

Powered by: