Service Independent Access Points (SIAP) to Optical Wide Area Networks
Joseph B. Evans, Victor S. Frost, Gary J. Minden

Information & Telecommunications Technology Center
University of Kansas
Lawrence, KS 66045

	In this talk, we describe architectures for high speed,
Service Independent Access Points (SIAPs) to high capacity (10 Gbps)
Optical Wide Area Networks (O-WANs).  The proposed SIAP architecture
is an advancement over the original GigaPop concept put forth in 1996.
The GigaPop was to emulate transparent networking services over
existing SONET based networks.  During the intervening eighteen months
the necessity to utilize the SONET infrastructure has been eliminated.
Today we propose operating service independent access directly to Wave
Division Multiplex (WDM) O-WANs.
	Optical networking systems technologies are constrained by the
attributes of the physical layer and components currently available.
Among these constraints are: (1) the relative difficulty of supporting
large address spaces in all-optical systems due to the cost and
complexity of components, (2) the relatively slow frame switching and
wavelength reorientation rates that can be obtained as compared to the
gross throughput, and (3) the lack of sophisticated optical logic
devices.  These constraints imply that optical TDM systems must use
large frames (compared to current technologies) and short labels or
addresses on those frames.  Our networking architecture is based upon
these constraints.
	The primary elements of a SIAP node are:
* Protocol Engine - Used to convert between existing networking services
  and the O-WAN transmission protocol.  This introduces direct support for
  multiple higher layer protocols at the bit transport layer. Bit sequences
  within the optical framing structure identify the particular service
  protocol engine (switch, router, etc.) to use.  This simplifies
  implementation and supports high speeds.  Direct demultiplexing to upper
  layer protocol engines avoids unnecessary processing and queuing delays.
* Optical Processor - Used to encode transmission frames, and select and
  decode received frames utilizing high capacity (10 Gbps) optical
  processing.  This provides the mechanism by which a flow is encapsulated
  and its associated framing fields are processed.  It supports datapath,
  control, and management functions such as framing, address/protocol
  identification, scrambling and descrambling, and error control.
* Link Quality Estimator - Used to estimate signal quality on each of
  many wavelengths (8-16 wavelengths) for OA&M functions.
* WDM Add/Drop Multiplexers - Used to select a single wavelength for
  further processing from a received WDM signal or inject a new
  wavelength into a transmitted signal.
* Protocol-specific engines - Equipment that transmits and receives
  SONET, ATM, IP or other networking service protocols.
	We are developing technologies for the implementation of the
first three of these elements.  The last two network elements are
available through commercial vendors and currently available equipment
is used in this project, for example, the Cienna 16 wavelength system,
Alcatel SONET ADMs, FORE/Nortel ATM switches, and Cisco 12000 series
routers available at the University of Kansas.