M X T 3 0 1 0
High-Performance Cell Processor
Typical Applications
Segmentation and Reassembly (SAR)
•
ATM/packet/TDM interworking
•
Service adaptation to cell-based
backplanes at up to 800Mbps
•
Combined SARing and cell-switching
in low port-density applications
ATM port processing
•
Traffic shaping and policing
WAN access “common card” services
using the MXT3020 coprocessor
•
AAL1, AAL2, AAL5
•
Inverse Multiplexing for ATM (IMA),
Cell Relay, Circuit Emulation Service,
T1/E1 UNI
Distinctive Characteristics
Leading-edge performance
•
Data rates up to 800Mbps
•
Bi-directional OC-3 or unidirectional OC-12
at full line rates with minimum packet size
•
Zero-time context switching
Complete programmability
The MXT3010 is the industry’s highest-performance programmable cell processing
engine. It offers the throughput normally associated with hardwired designs while
delivering all of the benefits of a programmable solution. Capable of supporting data
rates up to 800 Mbps, the MXT3010 can perform data path functions such as
segmentation/reassembly or traffic shaping for ATM interfaces running at up to
OC-12 speeds. With its efficient RISC processor, high-speed context switching,
and flexible cell scheduling system, the MXT3010 can easily integrate all current and
emerging ATM Forum traffic classes, including ABR, in a single application.
Developers can license prepackaged software products that run on the MXT3010
or may write their own custom software. MXT3010 software products include:
CellMaker
®
-622:
AAL5 SAR firmware supporting 1 x OC-12 or 4 x OC-3
configurations with integrated ABR traffic management.
CellMaker
®
-155:
AAL5 SAR firmware supporting 1 x OC-3 or lower data rates,
with integrated ABR traffic management.
CircuitMaker
™
:
firmware for ATM/TDM/packet interworking using the MXT3010
Cell Processor and MXT3020 Circuit Coprocessor.
LinkMaker
™
:
firmware for cost-effective Inverse Multiplexing for ATM.
•
Adapts to new or changing standards
via firmware upgrades
•
Supports custom features
Concurrent support of CBR, VBR-nrt,
VBR-rt, UBR, and ABR traffic classes
Full Suite of
Development Tools
M A K E R C O M M U N I C AT I O N S , I N C .
Applica
Segmentation and Reassembly
High Performance
The MXT3010 contains all of the necessary hardware support mechanisms to construct
a high-performance SAR. CRC-32 and CRC-10 calculations are performed in hardware
during data transfers. Independent DMA engines for each interface operate concurrently
to achieve high data throughput by dispatching and pipelining operations. The cell
scheduling system works with a combination of software algorithms and hardware-based
bandwidth-reservation and request-resolution mechanisms to support multiple classes
of service across up to 16,000 virtual connections.
Versatile
The MXT3010 adapts easily to both OC-3 and OC-12 rate SAR applications. An OC-3
SAR can be built using a single MXT3010 and consolidating user data and host control
traffic across the 1.6 Gbps Port1 interface in a shared memory subsystem. At OC-12, higher
data rates require specialization of the Port1 and Port2 interfaces as described in the caption
to the left. In addition, two MXT3010s are required, each supporting half-duplex operation,
to provide full-duplex OC-12 capability.
The MXT3010 is not limited to operation at only these data rates. With cell scheduling and
traffic shaping controlled by firmware, virtually any data rate up to 800 Mbps can be supported.
Customizable
Because a RISC core controls the MXT3010, custom cell sizes or formats can be
accommodated, as can proprietary traffic shaping algorithms and adaptation layers.
Excess CPU bandwidth, available in OC-3 applications, can be used for enhanced statistics,
protocol processing, or other host processor functions.
OC-3/OC-12 SAR.
The MXT3010 provides exceptional
traffic management in the form of per-VC
scheduling, the ability to support a variety
of custom traffic shaping algorithms, and
integrated RM cell processing for ABR.
The diagram below illustrates that at OC-
12 speeds, it is advantageous to separate
data traffic from host control traffic. This
way the high performance packet memory
can be optimized for the burst sizes and
rates common to segmentation and
reassembly, without incurring overhead
due to small, unpredictable command and
response messages.
ation
Exa
Single-point flow control.
The ability to combine ingress port
processing functions such as policing
and statistics gathering with egress
functions such as traffic shaping make
the MXT3010 an ideal device to
perform flow control. The MXT3010
can support a variety of traffic shaping
algorithms and can perform per-VC
scheduling for up to 16,000 VCs.
Traffic Shaping
Traffic Shaping relies on two fundamental mechanisms: per-virtual circuit queuing
and a Cell Scheduling System (CSS).
The MXT3010 simplifies the implementation of independent queues for tens of thousands of
virtual circuits with a powerful context maintenance model. Separate channel descriptors for
each VC provide the parameters necessary to track queue location and depth. The context
switching architecture of the MXT3010 ensures that the channel descriptors can be loaded,
manipulated and saved in the short cell interval time of OC-12 rates.
The chip performs scheduling through a combination of firmware algorithms running
on the SWAN core and a hardware bandwidth-reservation mechanism. By creating
service routines that are subsequently linked to by specific VCs, multiple separate traffic
shaping algorithms (leaky bucket VBR, ABR, etc.) can be co-resident in a port processing
application. Each VC activates its associated service routine during cell processing to
calculate the appropriate inter-cell interval for transmission bandwidth reservation.
These calculations can be complex in nature since they are handled by firmware running
on the SWAN core. Virtual Source/Virtual Destination ABR algorithms can be
implemented on the MXT3010. Actual transmission bandwidth is reserved through
calls dispatched by the SWAN to the Cell Scheduling System. Each request of the
CSS identifies a target transmission time. The CSS evaluates the current bandwidth
reservation state as maintained in a schedule table, and assigns the next appropriate
opportunity to the requesting VC. This assignment is recorded in the CSS table,
which ultimately drives the ordering of the VCs serviced for transmission.
Inverse Multiplexing for ATM (IMA)
IMA
involves splitting and sending a high bandwidth cell stream across multiple
lower bandwidth links, and then reestablishing the original cell ordering at the
receiver. It does this by inserting a framing and control protocol between the PHY
and ATM layers at both endpoints of the IMA connection. The IMA control protocol
cells contain information about IMA group status, IMA link status, loopback test
information, and link synchronization information.
An IMA engine needs to operate in the ATM cell stream at line rate. It must recognize
IMA control cells, extract those cells from the cell stream, process the cells, and based
on the cells’ content, adjust the IMA behavior.
amples
The MXT3010 provides an ideal platform for this engine. The programmability of the
device insures that an IMA system can evolve with changing standards. In addition
to the basic IMA function, the MXT3010 can provide advanced features such as traffic
shaping using per-VC queues and congestion management mechanisms such as early or
partial packet discard.
A complete Ethernet-to-ATM edge device can be built with the MXT3010 by incorporating
AAL5 SAR functions in the same application as the IMA protocol. In this manner,
Ethernet packets can be converted to ATM cells and transmitted over an IMA group
using a single MXT3010. A significant cost advantage is gained by combining these
functions into a single device.
Single Chip Solution.
The MXT3010 is unique in its ability
to perform both segmentation and
reassembly and IMA processing in
a single integrated solution.
Theory of
Operation
MXT3010 Features
SWAN™ Processor
•
Pipelined, single-cycle operations
•
Zero-time context switching
•
32-bit instruction set and 16-bit ALU
•
Modulo arithmetic and built-in
ALU branching
•
Scoreboarded register set
•
On-chip instruction cache
•
Operation at 100, 80, or 66 MHz
On-Chip Hardware Agents
The MXT3010 delivers exceptional system performance through the use of parallel hardware
agents, zero-time context switching, and an instruction set optimized for ATM cell
manipulation and transmission. The chip is built around the SWAN
TM
(Soft-Wired
ATM Network) processor, which is a specialized cell processing engine that supports
data rates of up to 800 Mbps by combining the pipelined architecture of a RISC
processor with the instruction set power of an ATM-specific CISC processor.
A number of hardware machines are integrated on-chip and operate in parallel to assist
the SWAN. Three fully independent DMA engines work concurrently to sustain high
throughput of data and control traffic through a high-speed multi-port internal cell
buffer RAM. The scheduling of this traffic is managed continuously by an internal Cell
Scheduling System (CSS).
The CSS includes a unique hardware scoreboard that can resolve scheduling conflicts
for up to 16,000 active connections. This deterministic behavior is essential when
maintaining independent traffic shaping parameters for each virtual connection in an
OC-12 link. For more detail on the CSS, see the traffic shaping application example
in this product brief.
External bus interfaces are available for each of the three DMA engines which access
the internal cell buffer RAM. The UTOPIA port connects to an ATM network through
a UTOPIA Level 2 multi-PHY interface. Port1 is a 32-bit DMA system interface which
runs a lightweight, burst-mode protocol designed to deliver very high throughput while
being easily adaptable to standard busses such as PCI or the i960. This port connects to
packet memory in SARing applications or to an external cell buffer in port processing
applications. Port2 is a 16-bit general-purpose interface, which supports both burst- and
non-burst-mode operations. This port can be used for host control messages, allowing Port1
memory to be optimized for handling network traffic, or for connecting to MXT3020
Circuit Coprocessors in TDM applications. An inter-chip communications subsystem
provides access to internal and external state information for application-specific use.
•
Three DMA engines
•
Cell Scheduling System
•
Cell buffer RAM
Cell Transmission
•
Per-VC scheduling
•
8 independent schedule tables
for ports or Virtual Paths
•
Software-controlled queuing,
with a library of queuing functions
ATM Adaption Layer
•
VPI/VCI reduction mechanism
•
CRC-32 generation and checking
•
CRC-10 generation and checking
ATM PHY Interface
•
UTOPIA Level 2 Multi-PHY
•
8-bit bidirectional or 16-bit unidirectional
•
Programmable HEC insertion
•
Clocked independently from other
system blocks
嵌入式系统
京公网安备 11010802033920号
Copyright © 2005-2026 EEWORLD.com.cn, Inc. All rights reserved
电子工程世界
