SDF and its "long-running public multitenant shell server" kin are all pretty old, and not one of them has been spared from slowdowns due to oversubscription.
It's sort of like an MMO: it's only fun when you're on a heavily-populated "shard", but that also means that you're all putting strain on the same server.
I'm surprised that there's no more modern implementation of the same multitenant "shell system" as a scale-free distributed architecture—like Mesos, z/OS, or any other "multi-tenant scheduler of interactive container-jobs with access to a common GlusterFS volume" kind of approach. A PaaS you can hang out in like a BBS.
Come to think, you could put one of these up on AWS, using EC2 autoscaling + K8s/Docker Swarm + Amazon EFS for storage, quite easily. Has anyone done this yet?
It's sort of like an MMO: it's only fun when you're on a heavily-populated "shard", but that also means that you're all putting strain on the same server.
I'm surprised that there's no more modern implementation of the same multitenant "shell system" as a scale-free distributed architecture—like Mesos, z/OS, or any other "multi-tenant scheduler of interactive container-jobs with access to a common GlusterFS volume" kind of approach. A PaaS you can hang out in like a BBS.
Come to think, you could put one of these up on AWS, using EC2 autoscaling + K8s/Docker Swarm + Amazon EFS for storage, quite easily. Has anyone done this yet?