When we decided to build a cloud that could run our customers’ most challenging enterprise workloads, we knew we had a big job ahead of us. We needed a fast and flat network, bare metal instances with the latest chipsets, and connectivity and storage capable of running massive I/O. When we learned at Supercomputing 2019 that we are one of only two public clouds in the list of the top 20 supercomputing architectures in the world put out by the Virtual Institute for IO in their IO500 benchmark, we knew we'd hit the mark.
Not only is Oracle Cloud included with the fastest compute architectures in the world, many of them on-premises deployments like research supercomputers, but our configuration costs much less than most others on the list. We can attain over 20 GiB/s in bandwidth and over 110,000 IOPS with only 30 client nodes running 480 processors—far fewer resources than some of the other configurations. Our new HPC file storage solution has a documented total cost (compute and storage) of just five cents per GB per month at scale. This is significantly lower than similar high-performance workloads running on other clouds or on premises.
We’re excited to push the limits of raw performance and to enable more individuals and organizations around the world to use our technology. Our customers get a huge amount of compute power and access to data on our cloud, and that’s powering solutions in the cloud that aren’t possible anywhere else.