IEEE Spectrum June, 2015 - 20

giving SuperComputerS
a SeCond Wind
laser links help old computers learn new tricks
The speed of high-performance added, causing mapping to break down,"
computing has soared from around says Koibuchi. Consequently, some appli100 gigaflops in 1993 to over 50 petaflops cations end up performing inefficiently-
today and is on course to hit the long- for instance, when their data is forced to
sought exascale (1018 floating-point opera- hop between a number of switches, or
tions per second) mark in the 2020s. Yet when an ordered parallel application
this remarkable supercomputing prog- is forced to adapt to a random topology
ress can be something of a super night- used for irregular applications.
Koibuchi and his colleagues from sevmare for the institutes and government
agencies asked to invest the hundreds eral universities believe that free-space
of millions, even billions of dollars that optics (FSO) can mitigate this breakdown
leading systems can cost.
"We are achieving a 1,000[-fold]
improvement over 10 years, so after
just 5 years a conventional supercomputer is no longer able to perform (at
the necessary standard) and has to be
trashed," says Michihiro Koibuchi, a
systems architect at Japan's National
Institute of Informatics, in Tokyo.
Koibuchi and his colleagues think
they have a solution that will let users
get more out of older machines: freespace optics, lasers that link supercom- and improve performance and resource
puter nodes through the air.
use. FSO uses laser light to transmit data
Typically, the thousands of processing through the air from one terminal to
nodes that make up a high-end supercom- another in its line of sight. With its wide,
puter are clustered into several popular gigabit-scale bandwidth, it is used outnetwork topologies depending on the com- doors in interbuilding links and in aeroputer's primary use. The topologies are space and satellite communications. So
implemented as cable-connected switches handling the 40-gigabit-per-second transthat link a select group of server-stuffed fer rates of high-performance computers
cabinets. With multiple users running indoors shouldn't be a problem.
different jobs at the same time, mapping
"Rather than relying on a few fixed
a program's communications needs to topologies, we propose using FSO termithe most suitable topology for a particu- nals mounted on the cabinets to provide
lar job is essential for efficient processing. line-of-sight communications between
But such mapping becomes problem- almost any two cabinets," says Koibuchi.
atic when the same system must serve "Network topologies can then be reconvery different types of applications at figured dynamically. This will help applithe same time.
cations better maintain their topologies,
"Day by day, fragmentation occurs as reduce fragmentation and latency, and
tasks end at different times and others are also cut the amount of cable used."
20

JUn 2015

nORTh aMERICan

SPECTRUM.IEEE.ORG

The idea gets support from Erik
D'Hollander, a parallel computing
expert at Ghent University, in Belgium.
"Because FSO terminals can be readily
adjusted for any type of topology, they
have a huge potential to replace the datacommunication fixed backbone between...
supercomputer cabinets," he says.
FSO terminals suitable for supercomputer use are not yet available, however,
so Koibuchi and his colleagues simulated
several terminal layouts in software. What
they found was a reduction in latency of
up to 9 percent and a 36 percent reduction in fiber-optic cable length. That's a
significant savings, considering that a system might have thousands of kilometers
of cables and they are discarded when
machines become obsolete.
To confirm these results physically, the
researchers constructed four small
prototype bidirectional FSO terminals from off-the-shelf components.
These included optical transceivers
rated at both 10 and 40 Gb/s, commodity infrared lasers, and collimator lenses used to direct the beam, as
well as motors and gears to aim the
terminals in any direction with a margin of error of just 0.003 degrees at
40 meters. At this distance, the tests
confirmed the simulation results and
achieved a data rate of 38 Gb/s.
Despite the tests' success, Koibuchi
admits that commercializing the technology is not likely to happen anytime soon.
"There is only one real hurdle left to overcome, but it's severe. At present there
are no mass-produced high-bandwidth
FSO terminals."
D'Hollander, too, sees the economics
of FSO as a problem. "I'm less optimistic
about the cost effectiveness of FSO terminals employing accurate steering and
tracking equipment," he says. "Perhaps
fixed FSO terminals without a steering
mechanism would be a viable alternative."
Koibuchi says his team is studying the
high-bandwidth FSO problem with NEC
Green Platforms Research Laboratories,
a unit of the Japanese maker of electronics and supercomputers. -john boyd

IstockpHoto

nEwS

http://SPECTRUM.IEEE.ORG

Table of Contents for the Digital Edition of IEEE Spectrum June, 2015