OPTICAL SWITCHING: All-optical processing reads packet headers
- To: <transhumantech@xxxxxxxxxxxxx>, <transhumantech@xxxxxxxxxxx>
- Subject: OPTICAL SWITCHING: All-optical processing reads packet headers
- From: Eugene Leitl <eugene.leitl@xxxxxxxxxxxxxxxxxxx>
- Date: Tue, 7 Mar 2000 17:55:23 -0800 (PST)
- Cc: <MISC>
(((Originally found by Brain Atkins of cooltech@
Of course, as to be expected, current protocol deficiencies will
now become apparent, as optical switches can only accomodate
low-complexity circuits)))
http://www.ragingbull.com/mboard/boards.cgi?board=STFZF&read=1686
Article Date:
December, 1999
Magazine Volume:
35
Issue:
12
OPTICAL SWITCHING: All-optical processing reads packet headers
Incorporating news from O plus E magazine, Tokyo
OSAKA-In what they say is the first demonstration of this transmission
technique, researchers have succeeded in sending data packets to a
destination specified in the address without having to convert the
optical signals to electrical signals. The researchers include Naoya
Wada at the Optical Communications Technologies Laboratory of the
Ministry of Communications and Posts, Kenichi Kitayama at the Osaka
University Graduate School Department of Engineering, and colleagues
Communication in venues such as the Internet involves dividing data
into smaller units called packets and sending them separately, with
destination information such as addresses included in a
header. Routers read the header information and send the packets along
the correct path. The data are restored at the destination (see
figure).
Current networks perform most of their functions, including routing,
electrically. Optics are used only in the transfer of information over
the length of the path. As the quantity of data transmitted over the
Internet steadily increases, it is clear that the electric
manipulations will eventually give rise to bottlenecks. All-optical
networks solve this problem because they allow optical signals to be
manipulated without being converted to electric signals. This
experiment has confirmed the proof of principle of an optical routing
component in such a network. The biggest challenge lies in creating a
header processor that performs without the conventional
light-to-electricity conversion.
In the experiment, address information was expressed in a code used in
optical code-division multiple access (OCDMA) technology-OCDMA is an
optical version of CDMA, which is a familiar technology in cellular
phones.
Multiple copies of a row of pulses are made inside the header
processor and are compared to the addresses in the table using optical
correlation techniques. When the addresses match, a sharp peak is
seen. This peak can be used to control an optical switch
array. Because the comparisons are done in parallel, the manipulations
dramatically shorten the process. A compact (a few square centimeters)
optical waveguide device, also developed by the researchers, is used
to compare the signals. This stable device is also used to generate
the optical code.
PHOTO. In an optical router, a packet enters from the left as an
optical signal, which is divided and sent to either an optical gate
switch array or a header processor. The header processor reads the
packet header, identifies the address, and turns on the optical
switches that route the information to the specified destination. The
header processor essentially contains a lookup table that matches the
address and destination.
During the test an 8-bit address was sent to its correct destination
in 35 ps (no data were attached because the experiment was
theoretical). This speed translates into processing 100 million
packets in one second-approximately 1000 times faster than electrical
routing.
Although this experiment focused on making address identification
optical, the group hopes to develop an integrated optical routing
device by transforming other components such as switch-control
components to an optical format.
Courtesy O plus E magazine, Tokyo
CORRECTION
In the October Japanwatch story "Homogeneous superfine silicon
particles light up" (p. 48), the name of one research institute was
incorrectly given as the Mechanical Research Laboratory. The accurate
English translation should read Mechanical Engineering Laboratory,
AIST-MITI.
http://lfw.pennwellnet.com/home/archivearticles.cfm?Section=Archives&ARTICLE_ID=54423&VERSION_NUM=1
Dick Milde