Conexant and Cisco sign deal to swap IP for chip sets
Conexant and Cisco sign deal to swap IP for chip sets
By Loring Wirbel, EE Times
October 18, 2000 (7:51 p.m. EST)
URL: http://www.eetimes.com/story/OEG20001018S0044
NEWPORT BEACH, Calif. Conexant Systems Inc. signed a pact with Cisco Systems Inc. this week that will give Conexant access to intellectual property related to Cisco's Dynamic Packet Transport (DPT) technology, while Cisco receives a supply of chip sets that support DPT. The work leading to the pact was critical in allowing Conexant to sample the CX29950 ring processor unit, a controller that supports the new Resilient Packet Ring architecture. Cisco's DPT technology is one proposal in front of a new IEEE study group on resilient rings. The study group is slated to become Working Group 802.17 at IEEE's 802 plenary meeting in November. Conexant is following Cisco's general model of providing protection switching at the routing layer but is calling its implementation the Spatial Reuse Protocol, since DPT refers to Cisco's specific implementation. The RPR study group entails many pr oposals for building counter-rotating fiber rings for metropolitan-area networks, borrowing concepts from Sonet, Gigabit Ethernet and the Fiber Distributed Data Interface. Cisco's DPT uses Layer 3 restoration, Nortel Networks is proposing a combined Layer 1/Layer 3 concept called InterWAN Packet Transport, Luminous Networks Inc. is including protection switching in a modified medium access control chip at Layer 2, and Lantern Communications Inc. is using Sonet Digital Wrapper technology. "We saw a lot of merit in the proposals from the smaller companies," said Lauren Schlicht, product line manager for broadband internetworking solutions at Conexant's Boulder, Colo., group. "But at the end of the day, we took the attitude that whatever Cisco implements will probably get adopted by the industry. "Still, the RPU architecture is flexible enough to follow where the IEEE study group goes." The concepts embodied in resilient packet rings are so new themselves that only a handful of equipment companies t hus far have taken up the RPR banner. Conexant is the first semiconductor company to make public its intent to provide RPR controllers. Conexant used a design for a ring-based packet controller, developed in the Boulder WAN operations and originally called the RAC-24. But in designing the CX29550, Conexant not only listened to Cisco's overall DPT strategy but consulted with other, new Conexant groups for an optimal design. Consulting operations included the Maker Communications unit, in Massachusetts, which will produce network coprocessors to be used with the RPU, and Israel-based Novanet Inc., which has sampled OC-48 CMOS Sonet framers that can operate in a Sonet-less fiber ring alongside the CX29550. Conexant program manager Vince Eberhard said RPRs allow nodes to be added to or deleted from a metro fiber ring without requiring that the ring be brought down or reconfigured. The RPR proposals share the philosophy that rings should not require a master controller, nor should they require a token, li ke token ring or FDDI. Any node on the network should help recover from fiber cuts or single-node failures by automatically initiating a ring wrap around the failure. The ring described its conditions to member nodes through the dynamic combination of control packets and data packets being sent. When a packet arrives at a node, it can be received, forwarded, received and forwarded for multicast traffic, stripped to check for error conditions, or examined for special cases such as ring wraps and pass-through modes. Every packet must be examined, so the ring operates totally in store-and-forward modes, with no cut-through operations. Some of the trickiest elements in getting such rings to operate efficiently are the need to implement fairness so that one node cannot hog all the bandwidth; the need to have guaranteed protection switching that can operate within Sonet's 50-ms constraints; and the need to perform automated topology discovery to recognize new nodes on the ring. Fairness is implemented through a series of transit buffers, which help carry out the Spatial Reuse Protocol at the heart of the Cisco approach. Ordinarily, Eberhard said, nodes must automatically set traffic-engineering constraints based on four traffic types: high-priority transit packets moving through a node, high-priority transmit packets being originated by a node, and low-priority versions of both transit and transmit packets. When a ring nears congestion, the nodes must perform both a global fairness check, to make sure the ring is operating optimally and fairly as a whole, and local optimization, to make sure that local subgroups are sharing bandwidth among themselves equitably. That is performed by having the low-priority transit buffer send congestion messages to other nodes on the ring. Intelligent Protection Switching (IPS) operates with constraints identical to those of Sonet's Automatic Protection Switching, but without reserving extra bandwidth, as Sonet requires. Auto-notification of rings is sent out as a Laye r 3 message from any RPR node. Finally, topology discovery messages are sent out on both fiber rings, and nodes respond by appending medium access control address information, using a 6-byte MAC field. The first generation of RPU from Conexant has dual OC-48 (2.5-Gbit/second) MACs, and Conexant is designing an OC-192 (10-Gbit) follow-on for release in 2001. The controller supports packet address lookup, multicasting, priority processing, fairness implementation, topology discovery, protection switching and rate limitation. The chip interfaces to external SRAM to implement an external transit buffer of either 500-kbyte or 1-Mbyte density. There are also packet-over-Sonet Physical Layer 3 interfaces to both external Sonet framers and network processors. "We have to continue to look at the partitioning of the controller over time," Schlicht said. "Do you throw the framer on-chip, or does that reduce flexibility? Are two MACs optimal? What's the best transit buffer size? We may want to offer differe nt devices over time, as packet rings are used more widely." Conexant also is leaving bets open as to what kinds of network equipment OEMs will leap on the RPR bandwagon. Some may be opportunistic companies that develop specialized nodes for existing Cisco networks, she said. Others will develop a range of dedicated RPR nodes, RPR/10-Gbit Ethernet nodes, RPR/Sonet nodes, or some other hybrid that has yet to be defined. Even a server could turn into an RPR node. And if Cisco's early success in cable TV hybrid fiber/coax networks is any indication, Schlicht said, the cable headend or wireless basestation could become a node for an RPR network, perhaps even earlier than traditional telco switches or routers.
Related News
- Conexant and Mindspeed Sign Multi-Year Licensing Deal With Sagantec
- BrainChip and Cisco Internet of Everything Innovation Center Sign Agreement to Demonstrate the Capabilities of BrainChip's Spiking Neural Adaptive Processor (SNAP) Technology
- Rambus Cryptography Research Division Licenses Security Technologies to Cisco
- Is Cisco Intel's $1 billion foundry deal?
- Conexant Adopts Target's IP Designer Tool-Suite to Build Next-Generation Foundation DSP IP
Breaking News
- Frontgrade Gaisler Unveils GR716B, a New Standard in Space-Grade Microcontrollers
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Cadence Unveils Arm-Based System Chiplet
Most Popular
- Cadence Unveils Arm-Based System Chiplet
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Esperanto Technologies and NEC Cooperate on Initiative to Advance Next Generation RISC-V Chips and Software Solutions for HPC
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- Arteris Selected by GigaDevice for Development in Next-Generation Automotive SoC With Enhanced FuSa Standards
E-mail This Article | Printer-Friendly Page |