When you look at the rapidly evolving infrastructure as a service (IaaS) market, you might get the impression that it’s a race to the bottom with commodity hardware, inexpensive storage options, ongoing price cuts and the like.
If a cloud service provider can’t differentiate itself, things can get very ugly. That’s why one of the key tenets of IBM Cloud is delivering unprecedented speed and performance. Automating and delivering raw computing power from bare metal servers with cutting-edge GPUs (graphics processing units) can’t just be commoditized.
An example of this raw performance can be found in recent benchmark tests we conducted with our friends at MapD and Bitfusion. Together, we were able to scale up to 64 Tesla K80 GPUs across 16 bare metal servers to filter, query and aggregate a 40-billion-row dataset in just 271 milliseconds. To put that in context: together with our partners, we can scan 147 billion rows per second.
The ability to more easily pool and scale GPUs spanning multiple nodes into a single system is a significant breakthrough, maybe even a game changer. It enables businesses to manage the most complex, compute-intensive workloads — from deep learning and cognitive to big data analytics — using an affordable, on-demand computing infrastructure. The ability to explore data and run queries in near real time from the IBM Cloud gives users “supercomputer”-like performance.
Think of IBM Cloud as the Bloodhound (the project that aims to build a car that can break the world land speed record) of IaaS platforms. In many ways, IBM is trying to set its own, new cloud speed record by building the world’s fastest and most scalable cloud platform. IBM is aiming to democratize access to supercomputing resources via the cloud. Companies in industries including financial services, telecommunications, retail and social media gain a real, competitive advantage in being able to run queries like the ones we’ve tested.
Stay tuned. IBM has more HPC news coming.
Share this post: