When you hear the word “Super Computer”, the first thing that comes into your mind is a huge & highly secured facility where a giant computer machine is processing the world’s most complex algorithms in just a few seconds. The world, however, has changed drastically in the last few years.
There was a time that supercomputers were accessible only to a small number of organisations, mostly government agencies as well as public research facilities and scientific organizations. The advent of cloud computing as well as the wide availability of advanced cloud-based workload control (CWM) tools have lowered the entry barriers substantially.
Last week YellowDog, a Bristol, United Kingdom-based Cloud workload management company, created a virtual supercomputer by using its proprietary platform. At its highest-performing capacity, albeit lasted for approximately 10 minutes only, it had created more than 3.2 million virtual CPUs.
Although it wasn’t comparable to Fugaku – the world’s fastest supercomputer – however, the performance was impressive enough to put it in the top 10 list of the world’s most powerful supercomputers at the very least for a short period of time.
A Quick Fire
The provisioning, done for the benefit of a pharmaceutical company, helped to run a well-known drug discovery software in one cluster. The calculation of the back of the envelope places the total cost of the project at around $65,000.
just to put things in perspective, this is equivalent to 33,333 instances of the AWS’s with 96 cores c5.24xlarge instance which is one of the instances that were used for the duration of the run (essentially identical to server baremetal as well as dedicated servers ) at $1.6013 per hour. This is $53,376 per hour, or $57,824 for the entirety of the run (65 minutes all in all).
“With access to this on-demand supercomputer, the researchers were able to analyze and screen 337 million compounds in 7 hours. To replicate that using their on-premises systems would have taken two months,” said Colin Bridger from AWS.
Milestone In Cloud Computing
The most remarkable aspect of it is the accessibility. It became immediately clear that going further this kind of power can be made accessible to anyone who is able to pay for it. It’s built on the same technology that powers the cloud computing ecosystem: web hosting, website builders, cloud storage, email services among many others.
CWM platforms have developed through time to create algorithms and machine-learning capabilities to determine the best source of computing regardless of origin or type.
For instance, one cloud provider could offer the lowest-cost spot compute, still, it may not be chosen by customers it won’t be available within the region specified by the customer, or if there weren’t enough servers of the specified type available with the same cloud provider. In this scenario, a different computing source is chosen.
Undoubtedly, such an outcome is encouraging enough and would prove to be a stepping stone to have more powerful and meaningful solutions in real life that require extreme computing power to provide a seamless and smooth experience to users.