What's new in rCUDA 20.07alpha?
- A completely new and disruptive internal architecture has been designed and implemented for the core of rCUDA. The new internal architecture is aimed at provide much better support to more CUDA applications and also better performance. Notice, however, that we have not checked yet the amount of applications supported neither their performance when using rCUDA
- A completely new communications layer has been implemented, which has been architected to provide much easier maintenance of the code and also
much better performance than previous versions of rCUDA
- Multi-tenancy is supported. That is, a real GPU can be virtualized into multiple GPUs, which can be concurrently provided to several applications
- The rCUDA server can simultaneously provide service across TCP and InfiniBand networks. That is, the rCUDA server can provide support to some applications by using TCP/IP at the same time that other applications are served using the InfiniBand network
- Support for functions in the Driver API has been noticeably improved
- The use of P2P data copies has been noticeably simplified, making it fully transparent to the user.
- GPU memory can be safely partitioned among different applications. In next releases of rCUDA we will disclose the public API to do so
- Next releases of rCUDA will include the rCUDA-smi tool, which is similar to the nvidia-smi tool except that remote GPUs are monitored
- Next releases of rCUDA will include the rCUDA GPU scheduler, intended to provide efficient integration of rCUDA with Slurm and other job schedulers
- Next releases of rCUDA will include the sbatch and srun commands required to integrate rCUDA with Slurm. Other job schedulers, such as PBSpro, could also be supported
The rCUDA Team hopes that you enjoy this new version of the rCUDA technology!