46 Terabits per second: this is the staggering bandwidth claimed by the latest version of the US network ESnet.
Today, most Internet connections can reach very comfortable speeds. Some operators even offer individual fiber packages that can reach 8 Gb per second under ideal conditions. But for the latest scientific laboratories, the situation is still very different.
ESnet proved it again recently. This is a network operated by the US Department of Energy (DoE); uses it to connect various facilities that include several supercomputers and various state-of-the-art laboratories spread across the land. They are connected by a large network that now includes more than 20,000 kilometers of fiber optic.
It is one of the most efficient networks on the planet. Each of its branches is capable of achieving extremely high flow rates; The DoE usually says it is the world’s fastest operational large-scale network.
And he has just further consolidated his position; with the arrival of the new version of the installation, christened ESnet6, any branch can now be transferred between 400 Gigabit and 1 Terabit per second. The network as a whole now supports a total bandwidth of 46 Terabits (ie 46,000 Gigabits) per second and forever!
Not a transfer speed record
However, an important distinction must be made: it is actually a bandwidth logging, not the raw data transfer rate. In the latter category, the record belongs to a facility at NCIT, a Japanese research center. They managed to achieve a turnover greater than 1 Petabit/sor more than one million Gigabits per second (see this ITEM).
The fundamental difference is that it is an experimental system which for the moment remains limited to research; in short it has nothing to do with ESnet6. The latter is already used daily by hundreds of researchers affiliated with US federal agencies. This makes this bandwidth record all the more impressive, as it is not an abstract proof of concept.
DoE staff also believe that the arrival of ESnet6 represents a major paradigm shift for applied research network architectures. According to HPCWorld and HPC wireAmerican engineers collaborated with AMD to implement an architecture that is flexible, scalable and largely automated.
“ Before Esnet6, each network element was treated as a set of pets says Inder Monda, director of science networks at Berkeley Lab. Now they are considered a ” real bunch with a real shepherd at the helm: a program called Orchestrator. It is responsible for rerouting large data streams dynamically while reconfiguring the network in real-time as needed.
Blessed bread for scholars
And these extraordinary performances will directly contribute to the work of researchers; ESnet6 will allow them to transfer, analyze and store staggering volumes of data. And you will need them, knowing the amount of data produced by simulators, telescopes, particle accelerators, genome sequencers and other modern equipment that exist today.
“ As the complexity of scientific instruments and the resolution of supercomputers increases, the scientific community faces a growing challenge explains Barbara Helland, deputy director of the DoE’s IT department.
” We have exponentially increasing amounts of data and need to transport, share and process it ever faster. “, she specifies. ” With Esnet6, DoE researchers are equipped with the most sophisticated technology available to help them address challenges such as climate, renewable energy, semiconductor manufacturing, quantum computing… Suffice to say, American researchers are already wringing their hands.