SciNet in collaboration with the University of Toronto has unveiled the most powerful supercomputer and christened it Niagara. Niagara comes to replace SciNet’s Tightly Coupled Cluster (TCS) and the aging General Purpose Cluster (GPC) which have been in use since 2009. Dr. Daniel Gruner, the tech chief at SciNet, said, “We run conventional data processing, analytics, simulations, and increasingly, machine learning as well. The capacity of the current system, coupled with the capability to scale out the full size of the cluster easily is vital for supporting our extensive community of scientists.”
In a video presentation, he continues to say, “There isn’t a single field of research that can happen without high-performance computing. All Canadians will benefit from Niagara since fundamental research is essential to the development of what we now refer to as applied research.” Niagara gets its name from the Niagara Falls located in Ontario Canada, and that symbolizes awe, discovery, and power.
The 3,780-node IBM iData cluster has been reduced to allow for Niagara to enter full production, which is expected to commence later this month. Niagara is available to researchers across the country dealing in astrophysics, oceanic research, artificial intelligence, climate change, and all other research fields requiring huge data.
Why Niagara was developed
Along with three other systems, Niagara was developed and deployed through the joint efforts of Compute Canada and a national association of supercomputing networks including the Calcul Quebec, ACENET, and WesGrid. These systems were designed with the aim of bringing the country’s ancient supercomputing systems into the petascale realm.
What’s impressive about the Niagara supercomputer?
Niagara comprises 1,500 condensed Lenovo ThinkSystem SD530 highly efficient nodes. Being 10 times more powerful than its predecessor, the high-performance nodes have 40 cores each. This provides a processing power of three petaflops and is supported by a storage capacity of 12 million gigabytes, an equivalent of 12 petabytes. The servers are connected by more than 40 kilometers of high-speed fiber optic cables. This is Niagara’s distinguishing trait, and it makes it possible to make use of all 60,000 cores simultaneously to solve a single problem.
The ThinkSystem SD530 came before Lenovo’s SD650, which is a warm-water cooled system. The latter is meant to power the soon to be launched SuperMUC-NG in Germany, at the Leibniz Supercomputing center. It has an advantage over the ThinkSystem SD530 of being able to operate the hottest and fastest Xeon Scalable Processors at full tilt. An example of this is the 205-watts Xeon Platinum 8180. Though the cooler running processors offer less performance, they are better in terms of pricing.
With the theoretical speed of around 4.61 petaflops along with a 3 petaflops Linpack Rmax, it is ranked among 10 of the world’s fastest supercomputers. Lenovo has deployed fastest computing systems to other countries such as Denmark, Norway, Spain, Australia, Italy and will deploy in Germany specifically in Leibniz Supercomputing Center. Supercomputing at Lenovo began once it acquired IBM’s x86 corporation for a ballpark figure of $2.1 billion in 2014.
Besides being powerful, Niagara needs less operational power as it uses a parallel system. This is the only one of its kind in Canada and is located in a nondescript and secure housing in Vaughn, Ontario.
What will be Niagara’s purpose?
Innovative research computing has become very common amongst scientists not only in Canada but the world over. Through its use, researchers will obtain the computing power needed to study and come up with solutions to the challenges that face the world today. Large-scale simulations necessary for genomics and other research disciplines will be availed.
Niagara will be used along with three other systems that are supposed to be deployed by Compute Canada host sites. The others include Arbutus, that is, the Arbutus which is built by Lenovo, with an OpenStack cloud system. There is also the Graham built by Huawei, and which will be located at the University of Waterloo, and has a speed of-of 2.6 petaflops, and the Cedar which boasts of a 3.7 petaflop, made by Dell system, and will be installed at the University of Waterloo.
Niagara faces the challenge of running a research code that will be meant to discover how the oceans in the world function, something that has never been done before. Richard Peltier of the University of Toronto will be the head of the team that deals with this. The work will largely depend on real-time data that will be streamed through sensors set at the oceans’ bottom. Peltier thinks that through running the code by the use of the entire system, ocean science will be studied more effectively through the discovery of new areas.
The Lenovo-powered supercomputer is designed to deliver and exascale-ready structure with unlimited scalability, which will accelerate innovation. This $18 million infrastructure was funded by the University of Toronto, the Government of Ontario, as well as the Canada Foundation for Innovation.
IT Xchange is a business partner of Lenovo providing the Options Continuation Program to VARs and end users using Lenovo technology in their IT infrastructure.