TNW-20-2_page_1
or "Jiickt The'National Security Agency’s review of emergi iil Globe at a Glance | Pointers | Spinouts
TNW-20-2_page_10
Predicting the performance of extreme-scale supercomputer networks Network topologies Supercomputers gain their performance edge from…
TNW-20-2_page_11
For example, if the leftmost node in the fat tree de- picted by figure 1(b) needed to communicate with the rightmost node, it could send…
TNW-20-2_page_12
Predicting the performance of extreme-scale supercomputer networks At each point in (virtual) time, the simulator con- siders the current…
TNW-20-2_page_13
The two mechanisms that underlie our approach are flow-based simulation and logical clocks. We now describe each of these in turn.…
TNW-20-2_page_14
Predicting the performance of extreme-scale supercomputer networks If node A sends data to node H via the route A- B-E-H (cyan links) at…
TNW-20-2_page_15
avoid erroneous readings, as discussed below). These would represent a costly but rarely useful expense. Furthermore, access to a single,…
TNW-20-2_page_16
Predicting the performance of extreme-scale supercomputer networks Logical time spent at various locations Event Alice’s House Carol’s…
TNW-20-2_page_17
both events). More subtly, the readings do not indicate which of Alice or Carol arrived first at the tea house (as both soccer > tea and…
TNW-20-2_page_18
Predicting the performance of extreme-scale supercomputer networks 6,000 Worse 5,000 4,000 3,000 Time (s) 2,000 1,000 Fat tree, Fat tree…
TNW-20-2_page_19
the 3-D torus requires 70% more time than the fat tree to transfer PAGOSA’ two billion messages. For both network topologies, quartering…
TNW-20-2_page_2
@ Next Wave The National Security Agency’s review of emerging technologies Editor’s column Managing Editor “End of the road for Roadrunner”…
TNW-20-2_page_20
Predicting the performance of extreme-scale supercomputer networks ens of thousands, which is what is typical for a more raditional network…
TNW-20-2_page_21
a member of Los Alamos National Laboratory’s Per- formance and Architecture team, involved in perfor- mance analysis of new large-scale…
TNW-20-2_page_22
David Prucnal, PE consequence of doing useful work with computers is the production of heat. Every watt of energy that goes into a computer…
TNW-20-2_page_23
The fundamental problem Before getting into the details of immersion cooling, let’s talk about the production of heat by computers and the…
TNW-20-2_page_24
Doing more with less: Cooling computers with oil pays off 22 How computers are cooled There are many different ways that computers are kept…
TNW-20-2_page_25
can absorb for a given temperature increase; it is measured in joules per degree Celsius (J/°C). > Specific heat capacity is the heat…
TNW-20-2_page_26
Doing more with less: Cooling computers with oil pays off 24 If there was a way to cool computers without moving exorbitant quantities of…
TNW-20-2_page_27
FIGURE 3. The immersion cooling system at the Laboratory for Physical Sciences, like the one pictured above, uses mineral oil to cool IT…
TNW-20-2_page_28
Doing more with less: Cooling computers with oil pays off 26 Oil-immersion systems also need to expel heat, and one way is through the use…
TNW-20-2_page_29
30% — Fan-Powered Air — Pump-Powered Oil 25% 20% 2 = 3 g 1% Power savings é due to server 10% fan removal 5% 0% 100% 120% 140% 160% 180%…
TNW-20-2_page_3
TNW-20-2 page 3
TNW-20-2_page_30
Doing more with less: Cooling computers with oil pays off 28 a small emulator is installed to trick the power supply into thinking the fan…
TNW-20-2_page_31
FEATURE designing, building, and optimizing high-reliability data centers. He joined the Agency as a power sys- tems engineer 10 years ago…
TNW-20-2_page_32
SP ie “Saves. Fer as a > ee Energy-efficient _ Superconducting computing rm _ coming up to spe d | Marc A. Manheimer ower and energy use by…
TNW-20-2_page_33
the TOP500 list is readily available [1, ive power demand of the TOP500 su- j rs Was about 0.25 gigawatts (GW) in 2012. Advanced Research…
TNW-20-2_page_34
Energy-efficient superconducting computing coming up to speed 32 FIGURE 1. The Jaguar XT5 supercomputer at Oak Ridge National Laboratory…
TNW-20-2_page_35
A brief introduction to quantum computing | Paul Lopata a_i 4 y p i. ~~ XX ba ie, ammutehe ae | > a f™&, | ‘ > s 'y em 7: j™ a a" pes Ae i…
TNW-20-2_page_36
Beyond digital: A brief introduction to quantum computing 34 Introduction Computers are based on logic. These fundamental rules of logic…
TNW-20-2_page_37
prime factors of a number—a problem of historic interest that gets extremely difficult as the number to be factored gets larger and larger.…
TNW-20-2_page_38
Beyond digital: A brief introduction to quantum computing operations before succumbing to its limited coherence time. (See the further…
TNW-20-2_page_39
Further reading Introduction to quantum computing Stephen Jordan at the National Institute FEATURE » Nielsen MA , Chuang IL. Quantum…
TNW-20-2_page_4
ing Ctrl + Shift. ene oneee
TNW-20-2_page_40
6“ ” 1 | Gene/Q 5 Blue Gene/Q GLOBE AT A GLANCE The Green500 provides a ranking of the most energy-efficient supercomputers in the world.…
TNW-20-2_page_41
GLOBE (pi BG/Q sn Gene/Q 2 Aurora Tigon 1 Eurora LEGEND Mflops/W Mega (i.e., million) floating-point operations per second per watt kW…
TNW-20-2_page_42
POINTERS bREEH SUPERLCOMPUTERS The Green500 announces the most energy-efficient supercomputers of June 2013 The Green500 list of June 2013…
TNW-20-2_page_43
3,208.83 Mflops/W 3,300 Megaflops per watt 490.20 600 Mflops/W 500 400 250.18 300 Mflops/W 200 = a 42.33 100 =— $< Mflops/W- 0 | — © @——@…
TNW-20-2_page_44
42 SPIN@UTS, News from the Technology Transfer Program Novel methods for manufacturing photonic logic devices raditional integrated…
TNW-20-2_page_45
FIGURE 1. An image of a fabricated mode transition- discrimination (MTD) photonic logic device with semiconductor laser edged facets and…
TNW-20-2_page_46
, | EN Ae. wy eee \ -, NATIONAL SECURITY AGENCY % 9 om CENTRAL SECURITY SERVICE a ed NC SRE 7, Sere B Defending Our Nation. Securing The…
TNW-20-2_page_5
modeling, simulation, il and emulation | Benjamin Payne, PhD, and Noel Wheeler Defining the future with driving force in modern society.…
TNW-20-2_page_6
Defining the future with modeling, simulation, and emulation Whatisa supercomputer? Processor [tinge 200+ gigaflops Interconnect 200+…
TNW-20-2_page_7
resilience, time-to-solution, and energy efficiency. These metrics are interdependent. For example, in- creasing the speed of an HPC system…
TNW-20-2_page_8
Defining the future with modeling, simulation, and emulation Simulator Core Services Power, Area cost projections | cotowaton | stn Vendor…
TNW-20-2_page_9
” Predicting the performance of extreme-scale supercomputer networks | Scott Pakin, Xin Yuan, and Michael Lang computers not much different…