There are two qualities of service available today using Lancium Compute.
Medium quality of service
Jobs submitted with medium quality of service are placed into a first-in-first-out (FIFO) queue. When a job gets to the head of the queue it is scheduled on the next resource that meets or exceeds the resource quality requested. If the job receives a better resource, it will not be charged for the better resource.
Low quality of service
Jobs submitted to the low quality of service queue only run when there are no medium priority jobs in the queue that can run on the unused resources. Low priority jobs will be pre-empted when a higher priority job arrives. If a job needs to be pre-empted, we select the job that has executed the least amount of time for preemption.
All jobs may be suspended or migrated at any time do to power availability. Jobs are NOT charged for time while they are suspended or migrating. Typical power suspensions are less than two hours.
During the beta test period, all Lancium Compute resources are free of charge.
Additional host RAM is available for one cent per hour per 4 GB.
CPU-only jobs can be scaled up to 16 cores. Pricing and RAM is linear in the number of CPUs
RAM — host memory available to your job.
VRAM — the total amount of VRAM available across the dedicated GPUs.
Tesla K80 cards contain 2 GPU cores per card