Broadcom Corporation

Avago Technologies International Sales Pte. Limited

Avago Technologies General IP (singapore) Pte. Ltd.

Tellabs San Jose, Inc.

Mark David Griswold

Mohammad Kamel Issa

Michael Hei-Lung Lau

Ryan E Hirth

Adam Julian Goldstein

Mitri Halabi

Vahid Tabatabaee

Amir Lehavot

Raghavan Menon

Chih-Yuan Chang

Niranjan Vaidya

Shaham Parvin

Xiaoyang Zheng

Mohan V Kalkunte

Lawrence Drew Davis

Michael Lau

Eugene Opsasnick

Mitchell Gordon McGee

Kandasamy Aravinthan

Nathan Shumway Hill

Mohammad Jafar Akhbarizadeh

A method for managing network traffic is shown. The method includes establishing a virtual tunnel between a source endpoint and a destination endpoint, the virtual tunnel including a plurality of data flow paths, each of the plurality of data flow streams connecting the source endpoint and the destination endpoint. The method includes providing, via the destination endpoint, a plurality of credits to the source endpoint, the plurality of credits provided via two or more of the plurality of data flow paths. The method includes updating, at the source endpoint, a data transmission sequence based on the received plurality of credits. The method includes providing a plurality of data packets based on the data transmission sequence to the destination endpoint.

Scalable E2E network architecture and components to support low latency and high throughput

A switch may operate in a cut-through mode and a store-and-forward mode. While in a default cut-through mode, the switch continuously monitors ports for certain health metrics. If those health metrics fall below a threshold, the switch changes to operate in a store-and-forward mode, either for a predetermined period of time or until the health metrics rise above a threshold, at which point the switch can resume cut-through mode operations. If health metrics fall below an even lower threshold, or remain below threshold for a predefined period of time, the switch can automatically alert a remote system or software process.

Cut-through latency and network fault diagnosis with limited error propagation

A system includes a pre-trigger buffer and a post-trigger buffer for recording entries related to specific network element. Buffer management monitoring circuitry captures entries leading up to a trigger criterion being met in the pre-trigger buffer and entries following the trigger criterion being met in the post-trigger buffer. The trigger criterion may include network element status, such as a threshold queue level; or an event, such as a dropped packet. The pre-trigger buffer may include a circular buffer in which older entries are overwritten by newer entries. Once the trigger condition is met, the pre-trigger buffer contents are held while the post-trigger buffer fills. Once the post-trigger buffer fills, the contents of the buffers may be read.

Buffer circuitry for monitoring network element status

Continuing to integrate more aggregate bandwidth and higher radix into switch devices is an economic imperative because it creates value both for the supplier and customer in large data center environments which are an increasingly important part of the marketplace. While new silicon processes continue to shrink transistor and other chip feature dimensions, process technology cannot be relied upon as a key driver of power reduction. Transitioning from 28 nm to 16 nm is a special case where FinFET provides additional power scaling, but subsequent FinFET nodes are not expected to deliver as substantial of power reductions to meet the desired increases in integration. The disclosed switch architecture attacks the power consumption problem by controlling the rate at which power-consuming activities occur.

Decoupled packet and data processing rates in switch devices

A network switch including a set of communication ports is provided. The communication ports may have an allocated prebuffer to store data during packet switching operations. The network switch may further include a calendar associated with the set of communication ports that provides bandwidth configuration for the set of communication ports. The network switch may further include a secondary calendar that may be dynamically setup. The secondary calendar may provide an alternative bandwidth configuration strategy for the set of communication ports. The switch includes circuitry that may increase the prebuffer size and upon the successful increase of the prebuffer size reconfigure the set of communication ports from the original calendar to the secondary calendar, without a reboot. The circuitry may reset the prebuffer size after reconfiguration is complete and the switch may continue operation according to the reconfigured settings.

System for flexible dynamic reassignment of throughput

Embodiments of the present disclosure provide efficient solutions for detecting and transmitting heartbeat frames at an optical line terminal (OLT) of an Ethernet passive optical network (EPON). The processing burden of a host CPU is reduced by shifting some of the processing responsibilities of the host CPU to an embedded CPU. An OAM capture FIFO buffer stores frames for later processing by the embedded CPU, and fields of an OAM vector register are changed whenever a heartbeat frame is detected for an LLID. Embedded CPU polls the OAM vector at a predetermined rate sufficient to maintain link status, so heartbeat messages are not missed even when the OAM capture FIFO buffer overflows. Additionally, an OAM generation buffer efficiently creates new heartbeat messages to be sent to ONUs for each supported LLID by modifying previously transmitted OAM frames with new information specific to the LLIDs.

OLT MAC module for efficiently processing OAM frames

A system and method for a flexible number of lookups in pipeline-based packet processors. Pipeline-based packet processors can be configured to allow multiple lookups per physical table. In one embodiment, a start of packet (SOP) cell is assigned to a first slot of a packet processing pipeline, and a non-SOP cell is assigned to a second slot of a packet processing pipeline. Access of a table by the second slot can be usurped by the SOP cell in the first slot.

System and method for a flexible number of lookups in pipeline-based packet processors

A device for instrumentation and traffic disposition of a network using one or more time-stamps may include a receiving port to receive a data packet. A device configuration module may be configured to determine whether the device is a boundary device located on a boundary of an instrumented sub-network of the network. If the determination is made that device is the boundary device, a frame processing module may insert a first time-stamp at a first offset from a frame checksum (FCS) field in a data frame associated with the data packet. Otherwise, a corresponding time-stamp may be inserted at a second offset from the FCS field. The one or more time-stamps may enable a receiving endpoint device of the network to determine timeliness information associated with the data packet.

Method and implementation for network instrumentation and traffic disposition with timestamps

A system for providing oversubscription of pipeline bandwidth comprises a steer module, an absorption buffer, an ingress packet processor (IPP), a memory management unit (MMU), and a main packet buffer. The steer module receives packets that include start of packet (SOP), middle of packet (MOP), and end of packet (EOP) cells, attaches a packet identifier to the cells, passes the MOP and EOP cells to the MMU, and stores the SOP cells and EOP metadata in the absorption buffer. The IPP processes the SOP cells and EOP metadata and passes the same to the MMU. The MMU stores the MOP, EOP, and processed SOP cells in the main packet buffer, combines, upon receiving the processed EOP metadata of each packet, the processed SOP cell, the MOP cells and the EOP cell of each packet to reconstruct each packet, and queues each reconstructed packet in an egress port queue.

Providing oversubscription of pipeline bandwidth

An optical line terminal (OLT) system for a passive optical network (PON) may include a processor, an OLT Medium Access Control (MAC) device communicatively coupled to PON ports, and a switch device communicatively coupled to the OLT MAC device via an Ethernet interface. The processor may map logical identifiers of each PON port to tunnel identifiers, where each tunnel identifier is indicative of a logical identifier and its corresponding PON port. The OLT MAC device may receive upstream data items that include logical identifiers over the PON ports. The OLT MAC device may replace the logical identifier of each upstream data item with the tunnel identifier that is mapped to the logical identifier and the PON port over which the upstream data item was received. The OLT MAC device may transmit, over the Ethernet interface to the switch device, the upstream data items including the tunnel identifiers.

Optical line terminal (OLT) system

A data rate control system for an optical line terminal (OLT) may include a processor, an OLT Medium Access Control (MAC) device that includes passive optical network (PON) ports that are mapped to identifiers, and a switch device coupled to the OLT MAC device. The OLT MAC device may determine that a PON port is congested, and may transmit a message to the switch device indicates the congestion at the PON port based on the mapped identifier. The switch device may transmit data items to the OLT MAC device for transmission over the PON ports at data rates respective to the PON ports, may receive the message, and may reduce the rate at which the data items are being transmitted to the OLT MAC device for transmission over the congested PON port without changing the rates at which other data items are being transmitted to the OLT MAC device.

Data rate control in an optical line terminal

Aspects of port empty transition scheduling are described herein. In one embodiment, when one or more cells are added to a queue in a network communications device, an enqueue indicator is generated. The enqueue indicator identifies a number of cells added to the queue. With reference to the enqueue indicator, a queue scheduler maintains a count of cells enqueued for communication and issues a port pick credit for a port of the network communications device. A port scheduler schedules a pick for communicating over the port with reference to the port pick credit and forwards the pick to the queue scheduler. In turn, the queue scheduler forwards a queue pick to the queue, and at least one of the cells is forwarded to dequeue logic. According to aspects of the embodiments described herein, empty port scheduling inefficiencies may be avoided and network throughput increased.

Port empty transition scheduling

Systems and methods of writing data to a buffer during a buffer cycle are described. The buffer has a plurality of buffer banks having various fill levels. The buffer determines a first portion of banks from the plurality of buffer banks. The first portion of banks unfilled banks. A rank can be assigned to each of the first portion of banks and a candidate set of banks chosen from the first portion of banks. A target bank is then chosen from the candidate set and the data is written to that bank. The ranking may be random. Furthermore, the target bank can be chosen based on ranking, fill level, or both.

Method and apparatus for probabilistic allocation in a switch packet buffer

Embodiments of the present invention relate to portions of a switch fabric having a single logical stage and at least one physical stage. In addition, the data paths and the control paths of the switch fabric can be decoupled thereby allowing additional processing to be performed than would otherwise be the case with control rates that matched the high data rates. In other words, data cells received on high speed links can be spread over many lower speed links; consequently, the data cells can transit the switch fabric at that high speed while the control information associated with the data can be processed at that lower speed. Because the control information can be processed at a lower speed (associated with the control path), the control information can be processed over a greater period of time.

Growing community of inventors

Fremont, CA, United States of America

Mark David Griswold