公开(公告)号：US11516133B2

公开(公告)日：2022-11-29

申请号：US16920993

申请日：2020-07-06

Applicant: Nokia Solutions and Networks OY

Inventor： Hyunseok Chang ,  Fang Hao ,  Muralidharan Kodialam ,  T. V. Lakshman ,  Sarit Mukherjee ,  Limin Wang

IPC: H04L12/747 ,  H04L12/721 ,  G06F12/0802 ,  H04L12/743 ,  H04L12/931 ,  H04L45/74 ,  H04L45/7453 ,  H04L49/00 ,  H04L45/00

Abstract: Packet-processing circuitry including one or more flow caches whose contents are managed using a cache-entry replacement policy that is implemented based on one or more updatable counters maintained for each of the cache entries. In an example embodiment, the implemented policy enables the flow cache to effectively catch and keep elephant flows by giving to the caught elephant flows appropriate preference in terms of the cache dwell time, which can beneficially improve the overall cache-hit ratio and/or packet-processing throughput. Some embodiments can be used to implement an Open Virtual Switch (OVS). Some embodiments are advantageously capable of implementing the cache-entry replacement policy with very limited additional memory allocation.

公开(公告)号：US11909656B1

公开(公告)日：2024-02-20

申请号：US18155416

申请日：2023-01-17

Applicant: Nokia Solutions and Networks Oy

Inventor： Hyunseok Chang ,  Sarit Mukherjee ,  Limin Wang

IPC: G06F15/16 ,  H04L47/2483 ,  H04L47/11 ,  H04L41/0895

CPC classification number: H04L47/2483 ,  H04L41/0895 ,  H04L47/11

Abstract: A CPU implements a virtual network function (VNF) associated with flow processing and can offload at least some of the flow processing to a hardware accelerator. A module (e.g., TOR switch), external to the software-based processor and the hardware accelerator, examines packet flows and determines (e.g., using a top-K elephant-flow-detection algorithm) which ones are more suitable for offloading from the software-based processor to the hardware accelerator than others. The external module communicates the relative suitability for offloading to the CPU using packet tags or out-of-band messaging, and the CPU uses the relative suitability to determine which flows to offload or onload. In some implementations, the CPU can instruct the external module to change the value of K used in the algorithm.

公开(公告)号：US20220006737A1

公开(公告)日：2022-01-06

申请号：US16920993

申请日：2020-07-06

Applicant: Nokia Solutions and Networks OY

Inventor： Hyunseok Chang ,  Fang Hao ,  Muralidharan Kodialam ,  T.V. Lakshman ,  Sarit Mukherjee ,  Limin Wang

IPC: H04L12/747 ,  G06F12/0802 ,  H04L12/721 ,  H04L12/743 ,  H04L12/931

Abstract: Packet-processing circuitry including one or more flow caches whose contents are managed using a cache-entry replacement policy that is implemented based on one or more updatable counters maintained for each of the cache entries. In an example embodiment, the implemented policy enables the flow cache to effectively catch and keep elephant flows by giving to the caught elephant flows appropriate preference in terms of the cache dwell time, which can beneficially improve the overall cache-hit ratio and/or packet-processing throughput. Some embodiments can be used to implement an Open Virtual Switch (OVS). Some embodiments are advantageously capable of implementing the cache-entry replacement policy with very limited additional memory allocation.

公开(公告)号：US11876691B2

公开(公告)日：2024-01-16

申请号：US17586969

申请日：2022-01-28

Applicant: Nokia Solutions and Networks Oy

Inventor： Hyunseok Chang ,  Limin Wang ,  Sarit Mukherjee ,  Walid Abdelrahman

IPC: H04L43/06 ,  G06F13/28 ,  H04L43/08

CPC classification number: H04L43/06 ,  G06F13/28 ,  H04L43/08 ,  G06F2213/28

Abstract: An end-to-end telemetry system for a Remote Direct Memory Access (RDMA) communication network having multiple end-servers. The system includes an RDMA tracer for each end-server, one or more programmable data planes, and a telemetry collector. Each RDMA tracer extracts host-level telemetry information for one or more RDMA sessions associated with the corresponding end-server. Each programmable data plane extracts network-level telemetry information for one or more RDMA sessions associated with the programmable data plane. The telemetry collector (i) receives the host-level telemetry information from the RDMA tracers and the network-level telemetry information from the one or more programmable data planes and (ii) generates telemetry reports based on the host-level and network-level telemetry information. In some implementations, the system enables real-time monitoring of RDMA traffic at the RDMA protocol level granularity across all RDMA-enabled workloads for different use cases.

公开(公告)号：US20230246931A1

公开(公告)日：2023-08-03

申请号：US17586969

申请日：2022-01-28

Applicant: Nokia Solutions and Networks Oy

Inventor： Hyunseok Chang ,  Limin Wang ,  Sarit Mukherjee ,  Walid Abdelrahman

IPC: H04L43/06 ,  H04L43/08 ,  G06F13/28

CPC classification number: H04L43/06 ,  H04L43/08 ,  G06F13/28 ,  G06F2213/28

Abstract: An end-to-end telemetry system for a Remote Direct Memory Access (RDMA) communication network having multiple end-servers. The system includes an RDMA tracer for each end-server, one or more programmable data planes, and a telemetry collector. Each RDMA tracer extracts host-level telemetry information for one or more RDMA sessions associated with the corresponding end-server. Each programmable data plane extracts network-level telemetry information for one or more RDMA sessions associated with the programmable data plane. The telemetry collector (i) receives the host-level telemetry information from the RDMA tracers and the network-level telemetry information from the one or more programmable data planes and (ii) generates telemetry reports based on the host-level and network-level telemetry information. In some implementations, the system enables real-time monitoring of RDMA traffic at the RDMA protocol level granularity across all RDMA-enabled workloads for different use cases.