Modern high-end computing systems including scientific
computing centers, cloud computing platforms and enterprise
computing systems use tens of thousands of nodes with
hundreds of thousands of cores. The upcoming 10-20 petaflop
machines such as the Livermore Sequoia, the Argonne Mira
supercomputers in the U.S., the Kei supercomputer in Japan,
would further increase this number to millions of cores.
Finally, with the upcoming exascale platforms, this number
is expected to further grow to millions of nodes with close to
a billion cores. Such massive-scale systems consume a large
amount of energy for operation and cooling. For example,
the 2.98 petaflop Dawning Nebula system requires close to
2.55 MW of power to operate. The increasing trend of power
consumption would keep growing, if there is no investment
into energy and power-awareness research for such systems
Modern high-end computing systems including scientific
computing centers, cloud computing platforms and enterprise
computing systems use tens of thousands of nodes with
hundreds of thousands of cores. The upcoming 10-20 petaflop
machines such as the Livermore Sequoia, the Argonne Mira
supercomputers in the U.S., the Kei supercomputer in Japan,
would further increase this number to millions of cores.
Finally, with the upcoming exascale platforms, this number
is expected to further grow to millions of nodes with close to
a billion cores. Such massive-scale systems consume a large
amount of energy for operation and cooling. For example,
the 2.98 petaflop Dawning Nebula system requires close to
2.55 MW of power to operate. The increasing trend of power
consumption would keep growing, if there is no investment
into energy and power-awareness research for such systems
การแปล กรุณารอสักครู่..