I've been placed in charge of getting Eclipse Che working with an online learning environment called STEM-CyLE. It's part of a NSF Research Grant to enhance Computer Science education with online learning engagement strategies and learning objects. I have a few questions regarding Che and how I can modify it to meet the requirements I have been given, in no way am I asking to make these changes to Che. I'm just in need of help and feel overwhelmed.
So I understand that the way Che works is that it provides a shared workspace for multiple developers to access. Contrary to how it works, the requirements that I have is to created a separate workspace for each individual user(student) that accesses Che, that is to say that student A should not be allowed to touch any files created by Student B. Can this be implemented? How? Where should I start looking?
Assuming I get the exclusive workspaces for Student to work, I also need it to scale with a finite amount of ram (about 72 GB, private server at a Uni with no intention of moving to the cloud due to FERPA regulations). So I was thinking that if each workspace instance would have its own JAVA jvm at most I could do would be about 30 students using Che at the same time, which is not realistic considering the platform is already being used at multiple institutions by about 500 students per semester. So my proposed solution was to somehow modify Che so that there is a queue that receives code to compile using a jvm. So instead of having one JVM running per student I have a set of 5 to begin with waiting for jobs(code to compile). As the queue fills up faster than it empties or if a specific jvm is taking longer than expected to be free for the next job, a new jvm will be started and added to the set in order to scale to the increased workload. Is this possible? If yes, where should I start looking?
My last requirement is that I make the authentication of the student into Che possible from the current authentication method which is through a JWT token that gets passed from client to backend on every request. Where should I look into so I can possibly create a solution for this?
--
Yours truly,
Hamilton Chevez
